2 * linux/fs/nfsd/nfs4state.c
4 * Copyright (c) 2001 The Regents of the University of Michigan.
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <kandros@umich.edu>
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/param.h>
38 #include <linux/major.h>
39 #include <linux/slab.h>
41 #include <linux/sunrpc/svc.h>
42 #include <linux/nfsd/nfsd.h>
43 #include <linux/nfsd/cache.h>
44 #include <linux/file.h>
45 #include <linux/mount.h>
46 #include <linux/workqueue.h>
47 #include <linux/smp_lock.h>
48 #include <linux/kthread.h>
49 #include <linux/nfs4.h>
50 #include <linux/nfsd/state.h>
51 #include <linux/nfsd/xdr4.h>
52 #include <linux/namei.h>
53 #include <linux/swap.h>
54 #include <linux/mutex.h>
55 #include <linux/lockd/bind.h>
56 #include <linux/module.h>
57 #include <linux/sunrpc/svcauth_gss.h>
59 #define NFSDDBG_FACILITY NFSDDBG_PROC
62 static time_t lease_time
= 90; /* default lease time */
63 static time_t user_lease_time
= 90;
64 static time_t boot_time
;
65 static u32 current_ownerid
= 1;
66 static u32 current_fileid
= 1;
67 static u32 current_delegid
= 1;
69 static stateid_t zerostateid
; /* bits all 0 */
70 static stateid_t onestateid
; /* bits all 1 */
72 #define ZERO_STATEID(stateid) (!memcmp((stateid), &zerostateid, sizeof(stateid_t)))
73 #define ONE_STATEID(stateid) (!memcmp((stateid), &onestateid, sizeof(stateid_t)))
75 /* forward declarations */
76 static struct nfs4_stateid
* find_stateid(stateid_t
*stid
, int flags
);
77 static struct nfs4_delegation
* find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
);
78 static char user_recovery_dirname
[PATH_MAX
] = "/var/lib/nfs/v4recovery";
79 static void nfs4_set_recdir(char *recdir
);
84 * protects clientid_hashtbl[], clientstr_hashtbl[],
85 * unconfstr_hashtbl[], uncofid_hashtbl[].
87 static DEFINE_MUTEX(client_mutex
);
89 static struct kmem_cache
*stateowner_slab
= NULL
;
90 static struct kmem_cache
*file_slab
= NULL
;
91 static struct kmem_cache
*stateid_slab
= NULL
;
92 static struct kmem_cache
*deleg_slab
= NULL
;
97 mutex_lock(&client_mutex
);
101 nfs4_unlock_state(void)
103 mutex_unlock(&client_mutex
);
107 opaque_hashval(const void *ptr
, int nbytes
)
109 unsigned char *cptr
= (unsigned char *) ptr
;
123 /* recall_lock protects the del_recall_lru */
124 static DEFINE_SPINLOCK(recall_lock
);
125 static struct list_head del_recall_lru
;
128 free_nfs4_file(struct kref
*kref
)
130 struct nfs4_file
*fp
= container_of(kref
, struct nfs4_file
, fi_ref
);
131 list_del(&fp
->fi_hash
);
133 kmem_cache_free(file_slab
, fp
);
137 put_nfs4_file(struct nfs4_file
*fi
)
139 kref_put(&fi
->fi_ref
, free_nfs4_file
);
143 get_nfs4_file(struct nfs4_file
*fi
)
145 kref_get(&fi
->fi_ref
);
148 static int num_delegations
;
149 unsigned int max_delegations
;
152 * Open owner state (share locks)
155 /* hash tables for nfs4_stateowner */
156 #define OWNER_HASH_BITS 8
157 #define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
158 #define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
160 #define ownerid_hashval(id) \
161 ((id) & OWNER_HASH_MASK)
162 #define ownerstr_hashval(clientid, ownername) \
163 (((clientid) + opaque_hashval((ownername.data), (ownername.len))) & OWNER_HASH_MASK)
165 static struct list_head ownerid_hashtbl
[OWNER_HASH_SIZE
];
166 static struct list_head ownerstr_hashtbl
[OWNER_HASH_SIZE
];
168 /* hash table for nfs4_file */
169 #define FILE_HASH_BITS 8
170 #define FILE_HASH_SIZE (1 << FILE_HASH_BITS)
171 #define FILE_HASH_MASK (FILE_HASH_SIZE - 1)
172 /* hash table for (open)nfs4_stateid */
173 #define STATEID_HASH_BITS 10
174 #define STATEID_HASH_SIZE (1 << STATEID_HASH_BITS)
175 #define STATEID_HASH_MASK (STATEID_HASH_SIZE - 1)
177 #define file_hashval(x) \
178 hash_ptr(x, FILE_HASH_BITS)
179 #define stateid_hashval(owner_id, file_id) \
180 (((owner_id) + (file_id)) & STATEID_HASH_MASK)
182 static struct list_head file_hashtbl
[FILE_HASH_SIZE
];
183 static struct list_head stateid_hashtbl
[STATEID_HASH_SIZE
];
185 static struct nfs4_delegation
*
186 alloc_init_deleg(struct nfs4_client
*clp
, struct nfs4_stateid
*stp
, struct svc_fh
*current_fh
, u32 type
)
188 struct nfs4_delegation
*dp
;
189 struct nfs4_file
*fp
= stp
->st_file
;
190 struct nfs4_callback
*cb
= &stp
->st_stateowner
->so_client
->cl_callback
;
192 dprintk("NFSD alloc_init_deleg\n");
193 if (fp
->fi_had_conflict
)
195 if (num_delegations
> max_delegations
)
197 dp
= kmem_cache_alloc(deleg_slab
, GFP_KERNEL
);
201 INIT_LIST_HEAD(&dp
->dl_perfile
);
202 INIT_LIST_HEAD(&dp
->dl_perclnt
);
203 INIT_LIST_HEAD(&dp
->dl_recall_lru
);
208 get_file(stp
->st_vfs_file
);
209 dp
->dl_vfs_file
= stp
->st_vfs_file
;
211 dp
->dl_recall
.cbr_dp
= NULL
;
212 dp
->dl_recall
.cbr_ident
= cb
->cb_ident
;
213 dp
->dl_recall
.cbr_trunc
= 0;
214 dp
->dl_stateid
.si_boot
= boot_time
;
215 dp
->dl_stateid
.si_stateownerid
= current_delegid
++;
216 dp
->dl_stateid
.si_fileid
= 0;
217 dp
->dl_stateid
.si_generation
= 0;
218 fh_copy_shallow(&dp
->dl_fh
, ¤t_fh
->fh_handle
);
220 atomic_set(&dp
->dl_count
, 1);
221 list_add(&dp
->dl_perfile
, &fp
->fi_delegations
);
222 list_add(&dp
->dl_perclnt
, &clp
->cl_delegations
);
227 nfs4_put_delegation(struct nfs4_delegation
*dp
)
229 if (atomic_dec_and_test(&dp
->dl_count
)) {
230 dprintk("NFSD: freeing dp %p\n",dp
);
231 put_nfs4_file(dp
->dl_file
);
232 kmem_cache_free(deleg_slab
, dp
);
237 /* Remove the associated file_lock first, then remove the delegation.
238 * lease_modify() is called to remove the FS_LEASE file_lock from
239 * the i_flock list, eventually calling nfsd's lock_manager
240 * fl_release_callback.
243 nfs4_close_delegation(struct nfs4_delegation
*dp
)
245 struct file
*filp
= dp
->dl_vfs_file
;
247 dprintk("NFSD: close_delegation dp %p\n",dp
);
248 dp
->dl_vfs_file
= NULL
;
249 /* The following nfsd_close may not actually close the file,
250 * but we want to remove the lease in any case. */
252 vfs_setlease(filp
, F_UNLCK
, &dp
->dl_flock
);
256 /* Called under the state lock. */
258 unhash_delegation(struct nfs4_delegation
*dp
)
260 list_del_init(&dp
->dl_perfile
);
261 list_del_init(&dp
->dl_perclnt
);
262 spin_lock(&recall_lock
);
263 list_del_init(&dp
->dl_recall_lru
);
264 spin_unlock(&recall_lock
);
265 nfs4_close_delegation(dp
);
266 nfs4_put_delegation(dp
);
273 /* Hash tables for nfs4_clientid state */
274 #define CLIENT_HASH_BITS 4
275 #define CLIENT_HASH_SIZE (1 << CLIENT_HASH_BITS)
276 #define CLIENT_HASH_MASK (CLIENT_HASH_SIZE - 1)
278 #define clientid_hashval(id) \
279 ((id) & CLIENT_HASH_MASK)
280 #define clientstr_hashval(name) \
281 (opaque_hashval((name), 8) & CLIENT_HASH_MASK)
283 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
284 * used in reboot/reset lease grace period processing
286 * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
287 * setclientid_confirmed info.
289 * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed
292 * client_lru holds client queue ordered by nfs4_client.cl_time
295 * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
296 * for last close replay.
298 static struct list_head reclaim_str_hashtbl
[CLIENT_HASH_SIZE
];
299 static int reclaim_str_hashtbl_size
= 0;
300 static struct list_head conf_id_hashtbl
[CLIENT_HASH_SIZE
];
301 static struct list_head conf_str_hashtbl
[CLIENT_HASH_SIZE
];
302 static struct list_head unconf_str_hashtbl
[CLIENT_HASH_SIZE
];
303 static struct list_head unconf_id_hashtbl
[CLIENT_HASH_SIZE
];
304 static struct list_head client_lru
;
305 static struct list_head close_lru
;
307 static void unhash_generic_stateid(struct nfs4_stateid
*stp
)
309 list_del(&stp
->st_hash
);
310 list_del(&stp
->st_perfile
);
311 list_del(&stp
->st_perstateowner
);
314 static void free_generic_stateid(struct nfs4_stateid
*stp
)
316 put_nfs4_file(stp
->st_file
);
317 kmem_cache_free(stateid_slab
, stp
);
320 static void release_lock_stateid(struct nfs4_stateid
*stp
)
322 unhash_generic_stateid(stp
);
323 locks_remove_posix(stp
->st_vfs_file
, (fl_owner_t
)stp
->st_stateowner
);
324 free_generic_stateid(stp
);
327 static void unhash_lockowner(struct nfs4_stateowner
*sop
)
329 struct nfs4_stateid
*stp
;
331 list_del(&sop
->so_idhash
);
332 list_del(&sop
->so_strhash
);
333 list_del(&sop
->so_perstateid
);
334 while (!list_empty(&sop
->so_stateids
)) {
335 stp
= list_first_entry(&sop
->so_stateids
,
336 struct nfs4_stateid
, st_perstateowner
);
337 release_lock_stateid(stp
);
341 static void release_lockowner(struct nfs4_stateowner
*sop
)
343 unhash_lockowner(sop
);
344 nfs4_put_stateowner(sop
);
348 release_stateid_lockowners(struct nfs4_stateid
*open_stp
)
350 struct nfs4_stateowner
*lock_sop
;
352 while (!list_empty(&open_stp
->st_lockowners
)) {
353 lock_sop
= list_entry(open_stp
->st_lockowners
.next
,
354 struct nfs4_stateowner
, so_perstateid
);
355 /* list_del(&open_stp->st_lockowners); */
356 BUG_ON(lock_sop
->so_is_open_owner
);
357 release_lockowner(lock_sop
);
361 static void release_open_stateid(struct nfs4_stateid
*stp
)
363 unhash_generic_stateid(stp
);
364 release_stateid_lockowners(stp
);
365 nfsd_close(stp
->st_vfs_file
);
366 free_generic_stateid(stp
);
369 static void unhash_openowner(struct nfs4_stateowner
*sop
)
371 struct nfs4_stateid
*stp
;
373 list_del(&sop
->so_idhash
);
374 list_del(&sop
->so_strhash
);
375 list_del(&sop
->so_perclient
);
376 list_del(&sop
->so_perstateid
); /* XXX: necessary? */
377 while (!list_empty(&sop
->so_stateids
)) {
378 stp
= list_first_entry(&sop
->so_stateids
,
379 struct nfs4_stateid
, st_perstateowner
);
380 release_open_stateid(stp
);
384 static void release_openowner(struct nfs4_stateowner
*sop
)
386 unhash_openowner(sop
);
387 list_del(&sop
->so_close_lru
);
388 nfs4_put_stateowner(sop
);
392 renew_client(struct nfs4_client
*clp
)
395 * Move client to the end to the LRU list.
397 dprintk("renewing client (clientid %08x/%08x)\n",
398 clp
->cl_clientid
.cl_boot
,
399 clp
->cl_clientid
.cl_id
);
400 list_move_tail(&clp
->cl_lru
, &client_lru
);
401 clp
->cl_time
= get_seconds();
404 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
406 STALE_CLIENTID(clientid_t
*clid
)
408 if (clid
->cl_boot
== boot_time
)
410 dprintk("NFSD stale clientid (%08x/%08x)\n",
411 clid
->cl_boot
, clid
->cl_id
);
416 * XXX Should we use a slab cache ?
417 * This type of memory management is somewhat inefficient, but we use it
418 * anyway since SETCLIENTID is not a common operation.
420 static struct nfs4_client
*alloc_client(struct xdr_netobj name
)
422 struct nfs4_client
*clp
;
424 clp
= kzalloc(sizeof(struct nfs4_client
), GFP_KERNEL
);
427 clp
->cl_name
.data
= kmalloc(name
.len
, GFP_KERNEL
);
428 if (clp
->cl_name
.data
== NULL
) {
432 memcpy(clp
->cl_name
.data
, name
.data
, name
.len
);
433 clp
->cl_name
.len
= name
.len
;
438 shutdown_callback_client(struct nfs4_client
*clp
)
440 struct rpc_clnt
*clnt
= clp
->cl_callback
.cb_client
;
444 * Callback threads take a reference on the client, so there
445 * should be no outstanding callbacks at this point.
447 clp
->cl_callback
.cb_client
= NULL
;
448 rpc_shutdown_client(clnt
);
453 free_client(struct nfs4_client
*clp
)
455 shutdown_callback_client(clp
);
456 if (clp
->cl_cred
.cr_group_info
)
457 put_group_info(clp
->cl_cred
.cr_group_info
);
458 kfree(clp
->cl_principal
);
459 kfree(clp
->cl_name
.data
);
464 put_nfs4_client(struct nfs4_client
*clp
)
466 if (atomic_dec_and_test(&clp
->cl_count
))
471 expire_client(struct nfs4_client
*clp
)
473 struct nfs4_stateowner
*sop
;
474 struct nfs4_delegation
*dp
;
475 struct list_head reaplist
;
477 dprintk("NFSD: expire_client cl_count %d\n",
478 atomic_read(&clp
->cl_count
));
480 INIT_LIST_HEAD(&reaplist
);
481 spin_lock(&recall_lock
);
482 while (!list_empty(&clp
->cl_delegations
)) {
483 dp
= list_entry(clp
->cl_delegations
.next
, struct nfs4_delegation
, dl_perclnt
);
484 dprintk("NFSD: expire client. dp %p, fp %p\n", dp
,
486 list_del_init(&dp
->dl_perclnt
);
487 list_move(&dp
->dl_recall_lru
, &reaplist
);
489 spin_unlock(&recall_lock
);
490 while (!list_empty(&reaplist
)) {
491 dp
= list_entry(reaplist
.next
, struct nfs4_delegation
, dl_recall_lru
);
492 list_del_init(&dp
->dl_recall_lru
);
493 unhash_delegation(dp
);
495 list_del(&clp
->cl_idhash
);
496 list_del(&clp
->cl_strhash
);
497 list_del(&clp
->cl_lru
);
498 while (!list_empty(&clp
->cl_openowners
)) {
499 sop
= list_entry(clp
->cl_openowners
.next
, struct nfs4_stateowner
, so_perclient
);
500 release_openowner(sop
);
502 put_nfs4_client(clp
);
505 static struct nfs4_client
*create_client(struct xdr_netobj name
, char *recdir
)
507 struct nfs4_client
*clp
;
509 clp
= alloc_client(name
);
512 memcpy(clp
->cl_recdir
, recdir
, HEXDIR_LEN
);
513 atomic_set(&clp
->cl_count
, 1);
514 atomic_set(&clp
->cl_callback
.cb_set
, 0);
515 INIT_LIST_HEAD(&clp
->cl_idhash
);
516 INIT_LIST_HEAD(&clp
->cl_strhash
);
517 INIT_LIST_HEAD(&clp
->cl_openowners
);
518 INIT_LIST_HEAD(&clp
->cl_delegations
);
519 INIT_LIST_HEAD(&clp
->cl_lru
);
523 static void copy_verf(struct nfs4_client
*target
, nfs4_verifier
*source
)
525 memcpy(target
->cl_verifier
.data
, source
->data
,
526 sizeof(target
->cl_verifier
.data
));
529 static void copy_clid(struct nfs4_client
*target
, struct nfs4_client
*source
)
531 target
->cl_clientid
.cl_boot
= source
->cl_clientid
.cl_boot
;
532 target
->cl_clientid
.cl_id
= source
->cl_clientid
.cl_id
;
535 static void copy_cred(struct svc_cred
*target
, struct svc_cred
*source
)
537 target
->cr_uid
= source
->cr_uid
;
538 target
->cr_gid
= source
->cr_gid
;
539 target
->cr_group_info
= source
->cr_group_info
;
540 get_group_info(target
->cr_group_info
);
543 static int same_name(const char *n1
, const char *n2
)
545 return 0 == memcmp(n1
, n2
, HEXDIR_LEN
);
549 same_verf(nfs4_verifier
*v1
, nfs4_verifier
*v2
)
551 return 0 == memcmp(v1
->data
, v2
->data
, sizeof(v1
->data
));
555 same_clid(clientid_t
*cl1
, clientid_t
*cl2
)
557 return (cl1
->cl_boot
== cl2
->cl_boot
) && (cl1
->cl_id
== cl2
->cl_id
);
560 /* XXX what about NGROUP */
562 same_creds(struct svc_cred
*cr1
, struct svc_cred
*cr2
)
564 return cr1
->cr_uid
== cr2
->cr_uid
;
567 static void gen_clid(struct nfs4_client
*clp
)
569 static u32 current_clientid
= 1;
571 clp
->cl_clientid
.cl_boot
= boot_time
;
572 clp
->cl_clientid
.cl_id
= current_clientid
++;
575 static void gen_confirm(struct nfs4_client
*clp
)
580 p
= (u32
*)clp
->cl_confirm
.data
;
581 *p
++ = get_seconds();
585 static int check_name(struct xdr_netobj name
)
589 if (name
.len
> NFS4_OPAQUE_LIMIT
) {
590 dprintk("NFSD: check_name: name too long(%d)!\n", name
.len
);
597 add_to_unconfirmed(struct nfs4_client
*clp
, unsigned int strhashval
)
599 unsigned int idhashval
;
601 list_add(&clp
->cl_strhash
, &unconf_str_hashtbl
[strhashval
]);
602 idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
603 list_add(&clp
->cl_idhash
, &unconf_id_hashtbl
[idhashval
]);
604 list_add_tail(&clp
->cl_lru
, &client_lru
);
605 clp
->cl_time
= get_seconds();
609 move_to_confirmed(struct nfs4_client
*clp
)
611 unsigned int idhashval
= clientid_hashval(clp
->cl_clientid
.cl_id
);
612 unsigned int strhashval
;
614 dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp
);
615 list_del_init(&clp
->cl_strhash
);
616 list_move(&clp
->cl_idhash
, &conf_id_hashtbl
[idhashval
]);
617 strhashval
= clientstr_hashval(clp
->cl_recdir
);
618 list_add(&clp
->cl_strhash
, &conf_str_hashtbl
[strhashval
]);
622 static struct nfs4_client
*
623 find_confirmed_client(clientid_t
*clid
)
625 struct nfs4_client
*clp
;
626 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
628 list_for_each_entry(clp
, &conf_id_hashtbl
[idhashval
], cl_idhash
) {
629 if (same_clid(&clp
->cl_clientid
, clid
))
635 static struct nfs4_client
*
636 find_unconfirmed_client(clientid_t
*clid
)
638 struct nfs4_client
*clp
;
639 unsigned int idhashval
= clientid_hashval(clid
->cl_id
);
641 list_for_each_entry(clp
, &unconf_id_hashtbl
[idhashval
], cl_idhash
) {
642 if (same_clid(&clp
->cl_clientid
, clid
))
648 static struct nfs4_client
*
649 find_confirmed_client_by_str(const char *dname
, unsigned int hashval
)
651 struct nfs4_client
*clp
;
653 list_for_each_entry(clp
, &conf_str_hashtbl
[hashval
], cl_strhash
) {
654 if (same_name(clp
->cl_recdir
, dname
))
660 static struct nfs4_client
*
661 find_unconfirmed_client_by_str(const char *dname
, unsigned int hashval
)
663 struct nfs4_client
*clp
;
665 list_for_each_entry(clp
, &unconf_str_hashtbl
[hashval
], cl_strhash
) {
666 if (same_name(clp
->cl_recdir
, dname
))
672 /* a helper function for parse_callback */
674 parse_octet(unsigned int *lenp
, char **addrp
)
676 unsigned int len
= *lenp
;
688 if ((c
< '0') || (c
> '9')) {
694 n
= (n
* 10) + (c
- '0');
705 /* parse and set the setclientid ipv4 callback address */
707 parse_ipv4(unsigned int addr_len
, char *addr_val
, unsigned int *cbaddrp
, unsigned short *cbportp
)
712 u32 addrlen
= addr_len
;
713 char *addr
= addr_val
;
718 for(i
= 4; i
> 0 ; i
--) {
719 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
722 cbaddr
|= (temp
<< shift
);
730 for(i
= 2; i
> 0 ; i
--) {
731 if ((temp
= parse_octet(&addrlen
, &addr
)) < 0) {
734 cbport
|= (temp
<< shift
);
743 gen_callback(struct nfs4_client
*clp
, struct nfsd4_setclientid
*se
)
745 struct nfs4_callback
*cb
= &clp
->cl_callback
;
747 /* Currently, we only support tcp for the callback channel */
748 if ((se
->se_callback_netid_len
!= 3) || memcmp((char *)se
->se_callback_netid_val
, "tcp", 3))
751 if ( !(parse_ipv4(se
->se_callback_addr_len
, se
->se_callback_addr_val
,
752 &cb
->cb_addr
, &cb
->cb_port
)))
754 cb
->cb_prog
= se
->se_callback_prog
;
755 cb
->cb_ident
= se
->se_callback_ident
;
758 dprintk(KERN_INFO
"NFSD: this client (clientid %08x/%08x) "
759 "will not receive delegations\n",
760 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
766 nfsd4_setclientid(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
767 struct nfsd4_setclientid
*setclid
)
769 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
770 struct xdr_netobj clname
= {
771 .len
= setclid
->se_namelen
,
772 .data
= setclid
->se_name
,
774 nfs4_verifier clverifier
= setclid
->se_verf
;
775 unsigned int strhashval
;
776 struct nfs4_client
*conf
, *unconf
, *new;
779 char dname
[HEXDIR_LEN
];
781 if (!check_name(clname
))
784 status
= nfs4_make_rec_clidname(dname
, &clname
);
789 * XXX The Duplicate Request Cache (DRC) has been checked (??)
790 * We get here on a DRC miss.
793 strhashval
= clientstr_hashval(dname
);
796 conf
= find_confirmed_client_by_str(dname
, strhashval
);
798 /* RFC 3530 14.2.33 CASE 0: */
799 status
= nfserr_clid_inuse
;
800 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
)
801 || conf
->cl_addr
!= sin
->sin_addr
.s_addr
) {
802 dprintk("NFSD: setclientid: string in use by clientat %pI4\n",
808 * section 14.2.33 of RFC 3530 (under the heading "IMPLEMENTATION")
809 * has a description of SETCLIENTID request processing consisting
810 * of 5 bullet points, labeled as CASE0 - CASE4 below.
812 unconf
= find_unconfirmed_client_by_str(dname
, strhashval
);
813 status
= nfserr_resource
;
816 * RFC 3530 14.2.33 CASE 4:
817 * placed first, because it is the normal case
820 expire_client(unconf
);
821 new = create_client(clname
, dname
);
825 } else if (same_verf(&conf
->cl_verifier
, &clverifier
)) {
827 * RFC 3530 14.2.33 CASE 1:
828 * probable callback update
831 /* Note this is removing unconfirmed {*x***},
832 * which is stronger than RFC recommended {vxc**}.
833 * This has the advantage that there is at most
834 * one {*x***} in either list at any time.
836 expire_client(unconf
);
838 new = create_client(clname
, dname
);
841 copy_clid(new, conf
);
842 } else if (!unconf
) {
844 * RFC 3530 14.2.33 CASE 2:
845 * probable client reboot; state will be removed if
848 new = create_client(clname
, dname
);
854 * RFC 3530 14.2.33 CASE 3:
855 * probable client reboot; state will be removed if
858 expire_client(unconf
);
859 new = create_client(clname
, dname
);
864 copy_verf(new, &clverifier
);
865 new->cl_addr
= sin
->sin_addr
.s_addr
;
866 new->cl_flavor
= rqstp
->rq_flavor
;
867 princ
= svc_gss_principal(rqstp
);
869 new->cl_principal
= kstrdup(princ
, GFP_KERNEL
);
870 if (new->cl_principal
== NULL
) {
875 copy_cred(&new->cl_cred
, &rqstp
->rq_cred
);
877 gen_callback(new, setclid
);
878 add_to_unconfirmed(new, strhashval
);
879 setclid
->se_clientid
.cl_boot
= new->cl_clientid
.cl_boot
;
880 setclid
->se_clientid
.cl_id
= new->cl_clientid
.cl_id
;
881 memcpy(setclid
->se_confirm
.data
, new->cl_confirm
.data
, sizeof(setclid
->se_confirm
.data
));
890 * Section 14.2.34 of RFC 3530 (under the heading "IMPLEMENTATION") has
891 * a description of SETCLIENTID_CONFIRM request processing consisting of 4
892 * bullets, labeled as CASE1 - CASE4 below.
895 nfsd4_setclientid_confirm(struct svc_rqst
*rqstp
,
896 struct nfsd4_compound_state
*cstate
,
897 struct nfsd4_setclientid_confirm
*setclientid_confirm
)
899 struct sockaddr_in
*sin
= svc_addr_in(rqstp
);
900 struct nfs4_client
*conf
, *unconf
;
901 nfs4_verifier confirm
= setclientid_confirm
->sc_confirm
;
902 clientid_t
* clid
= &setclientid_confirm
->sc_clientid
;
905 if (STALE_CLIENTID(clid
))
906 return nfserr_stale_clientid
;
908 * XXX The Duplicate Request Cache (DRC) has been checked (??)
909 * We get here on a DRC miss.
914 conf
= find_confirmed_client(clid
);
915 unconf
= find_unconfirmed_client(clid
);
917 status
= nfserr_clid_inuse
;
918 if (conf
&& conf
->cl_addr
!= sin
->sin_addr
.s_addr
)
920 if (unconf
&& unconf
->cl_addr
!= sin
->sin_addr
.s_addr
)
924 * section 14.2.34 of RFC 3530 has a description of
925 * SETCLIENTID_CONFIRM request processing consisting
926 * of 4 bullet points, labeled as CASE1 - CASE4 below.
928 if (conf
&& unconf
&& same_verf(&confirm
, &unconf
->cl_confirm
)) {
930 * RFC 3530 14.2.34 CASE 1:
933 if (!same_creds(&conf
->cl_cred
, &unconf
->cl_cred
))
934 status
= nfserr_clid_inuse
;
936 /* XXX: We just turn off callbacks until we can handle
937 * change request correctly. */
938 atomic_set(&conf
->cl_callback
.cb_set
, 0);
940 nfsd4_remove_clid_dir(unconf
);
941 expire_client(unconf
);
945 } else if (conf
&& !unconf
) {
947 * RFC 3530 14.2.34 CASE 2:
948 * probable retransmitted request; play it safe and
951 if (!same_creds(&conf
->cl_cred
, &rqstp
->rq_cred
))
952 status
= nfserr_clid_inuse
;
955 } else if (!conf
&& unconf
956 && same_verf(&unconf
->cl_confirm
, &confirm
)) {
958 * RFC 3530 14.2.34 CASE 3:
959 * Normal case; new or rebooted client:
961 if (!same_creds(&unconf
->cl_cred
, &rqstp
->rq_cred
)) {
962 status
= nfserr_clid_inuse
;
965 clientstr_hashval(unconf
->cl_recdir
);
966 conf
= find_confirmed_client_by_str(unconf
->cl_recdir
,
969 nfsd4_remove_clid_dir(conf
);
972 move_to_confirmed(unconf
);
974 nfsd4_probe_callback(conf
);
977 } else if ((!conf
|| (conf
&& !same_verf(&conf
->cl_confirm
, &confirm
)))
978 && (!unconf
|| (unconf
&& !same_verf(&unconf
->cl_confirm
,
981 * RFC 3530 14.2.34 CASE 4:
982 * Client probably hasn't noticed that we rebooted yet.
984 status
= nfserr_stale_clientid
;
986 /* check that we have hit one of the cases...*/
987 status
= nfserr_clid_inuse
;
994 /* OPEN Share state helper functions */
995 static inline struct nfs4_file
*
996 alloc_init_file(struct inode
*ino
)
998 struct nfs4_file
*fp
;
999 unsigned int hashval
= file_hashval(ino
);
1001 fp
= kmem_cache_alloc(file_slab
, GFP_KERNEL
);
1003 kref_init(&fp
->fi_ref
);
1004 INIT_LIST_HEAD(&fp
->fi_hash
);
1005 INIT_LIST_HEAD(&fp
->fi_stateids
);
1006 INIT_LIST_HEAD(&fp
->fi_delegations
);
1007 list_add(&fp
->fi_hash
, &file_hashtbl
[hashval
]);
1008 fp
->fi_inode
= igrab(ino
);
1009 fp
->fi_id
= current_fileid
++;
1010 fp
->fi_had_conflict
= false;
1017 nfsd4_free_slab(struct kmem_cache
**slab
)
1021 kmem_cache_destroy(*slab
);
1026 nfsd4_free_slabs(void)
1028 nfsd4_free_slab(&stateowner_slab
);
1029 nfsd4_free_slab(&file_slab
);
1030 nfsd4_free_slab(&stateid_slab
);
1031 nfsd4_free_slab(&deleg_slab
);
1035 nfsd4_init_slabs(void)
1037 stateowner_slab
= kmem_cache_create("nfsd4_stateowners",
1038 sizeof(struct nfs4_stateowner
), 0, 0, NULL
);
1039 if (stateowner_slab
== NULL
)
1041 file_slab
= kmem_cache_create("nfsd4_files",
1042 sizeof(struct nfs4_file
), 0, 0, NULL
);
1043 if (file_slab
== NULL
)
1045 stateid_slab
= kmem_cache_create("nfsd4_stateids",
1046 sizeof(struct nfs4_stateid
), 0, 0, NULL
);
1047 if (stateid_slab
== NULL
)
1049 deleg_slab
= kmem_cache_create("nfsd4_delegations",
1050 sizeof(struct nfs4_delegation
), 0, 0, NULL
);
1051 if (deleg_slab
== NULL
)
1056 dprintk("nfsd4: out of memory while initializing nfsv4\n");
1061 nfs4_free_stateowner(struct kref
*kref
)
1063 struct nfs4_stateowner
*sop
=
1064 container_of(kref
, struct nfs4_stateowner
, so_ref
);
1065 kfree(sop
->so_owner
.data
);
1066 kmem_cache_free(stateowner_slab
, sop
);
1069 static inline struct nfs4_stateowner
*
1070 alloc_stateowner(struct xdr_netobj
*owner
)
1072 struct nfs4_stateowner
*sop
;
1074 if ((sop
= kmem_cache_alloc(stateowner_slab
, GFP_KERNEL
))) {
1075 if ((sop
->so_owner
.data
= kmalloc(owner
->len
, GFP_KERNEL
))) {
1076 memcpy(sop
->so_owner
.data
, owner
->data
, owner
->len
);
1077 sop
->so_owner
.len
= owner
->len
;
1078 kref_init(&sop
->so_ref
);
1081 kmem_cache_free(stateowner_slab
, sop
);
1086 static struct nfs4_stateowner
*
1087 alloc_init_open_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfsd4_open
*open
) {
1088 struct nfs4_stateowner
*sop
;
1089 struct nfs4_replay
*rp
;
1090 unsigned int idhashval
;
1092 if (!(sop
= alloc_stateowner(&open
->op_owner
)))
1094 idhashval
= ownerid_hashval(current_ownerid
);
1095 INIT_LIST_HEAD(&sop
->so_idhash
);
1096 INIT_LIST_HEAD(&sop
->so_strhash
);
1097 INIT_LIST_HEAD(&sop
->so_perclient
);
1098 INIT_LIST_HEAD(&sop
->so_stateids
);
1099 INIT_LIST_HEAD(&sop
->so_perstateid
); /* not used */
1100 INIT_LIST_HEAD(&sop
->so_close_lru
);
1102 list_add(&sop
->so_idhash
, &ownerid_hashtbl
[idhashval
]);
1103 list_add(&sop
->so_strhash
, &ownerstr_hashtbl
[strhashval
]);
1104 list_add(&sop
->so_perclient
, &clp
->cl_openowners
);
1105 sop
->so_is_open_owner
= 1;
1106 sop
->so_id
= current_ownerid
++;
1107 sop
->so_client
= clp
;
1108 sop
->so_seqid
= open
->op_seqid
;
1109 sop
->so_confirmed
= 0;
1110 rp
= &sop
->so_replay
;
1111 rp
->rp_status
= nfserr_serverfault
;
1113 rp
->rp_buf
= rp
->rp_ibuf
;
1118 init_stateid(struct nfs4_stateid
*stp
, struct nfs4_file
*fp
, struct nfsd4_open
*open
) {
1119 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1120 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
1122 INIT_LIST_HEAD(&stp
->st_hash
);
1123 INIT_LIST_HEAD(&stp
->st_perstateowner
);
1124 INIT_LIST_HEAD(&stp
->st_lockowners
);
1125 INIT_LIST_HEAD(&stp
->st_perfile
);
1126 list_add(&stp
->st_hash
, &stateid_hashtbl
[hashval
]);
1127 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
1128 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
1129 stp
->st_stateowner
= sop
;
1132 stp
->st_stateid
.si_boot
= boot_time
;
1133 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
1134 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
1135 stp
->st_stateid
.si_generation
= 0;
1136 stp
->st_access_bmap
= 0;
1137 stp
->st_deny_bmap
= 0;
1138 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
1139 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1140 stp
->st_openstp
= NULL
;
1144 move_to_close_lru(struct nfs4_stateowner
*sop
)
1146 dprintk("NFSD: move_to_close_lru nfs4_stateowner %p\n", sop
);
1148 list_move_tail(&sop
->so_close_lru
, &close_lru
);
1149 sop
->so_time
= get_seconds();
1153 same_owner_str(struct nfs4_stateowner
*sop
, struct xdr_netobj
*owner
,
1156 return (sop
->so_owner
.len
== owner
->len
) &&
1157 0 == memcmp(sop
->so_owner
.data
, owner
->data
, owner
->len
) &&
1158 (sop
->so_client
->cl_clientid
.cl_id
== clid
->cl_id
);
1161 static struct nfs4_stateowner
*
1162 find_openstateowner_str(unsigned int hashval
, struct nfsd4_open
*open
)
1164 struct nfs4_stateowner
*so
= NULL
;
1166 list_for_each_entry(so
, &ownerstr_hashtbl
[hashval
], so_strhash
) {
1167 if (same_owner_str(so
, &open
->op_owner
, &open
->op_clientid
))
1173 /* search file_hashtbl[] for file */
1174 static struct nfs4_file
*
1175 find_file(struct inode
*ino
)
1177 unsigned int hashval
= file_hashval(ino
);
1178 struct nfs4_file
*fp
;
1180 list_for_each_entry(fp
, &file_hashtbl
[hashval
], fi_hash
) {
1181 if (fp
->fi_inode
== ino
) {
1189 static inline int access_valid(u32 x
)
1191 if (x
< NFS4_SHARE_ACCESS_READ
)
1193 if (x
> NFS4_SHARE_ACCESS_BOTH
)
1198 static inline int deny_valid(u32 x
)
1200 /* Note: unlike access bits, deny bits may be zero. */
1201 return x
<= NFS4_SHARE_DENY_BOTH
;
1205 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1206 * st_{access,deny}_bmap field of the stateid, in order to track not
1207 * only what share bits are currently in force, but also what
1208 * combinations of share bits previous opens have used. This allows us
1209 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1210 * return an error if the client attempt to downgrade to a combination
1211 * of share bits not explicable by closing some of its previous opens.
1213 * XXX: This enforcement is actually incomplete, since we don't keep
1214 * track of access/deny bit combinations; so, e.g., we allow:
1216 * OPEN allow read, deny write
1217 * OPEN allow both, deny none
1218 * DOWNGRADE allow read, deny none
1220 * which we should reject.
1223 set_access(unsigned int *access
, unsigned long bmap
) {
1227 for (i
= 1; i
< 4; i
++) {
1228 if (test_bit(i
, &bmap
))
1234 set_deny(unsigned int *deny
, unsigned long bmap
) {
1238 for (i
= 0; i
< 4; i
++) {
1239 if (test_bit(i
, &bmap
))
1245 test_share(struct nfs4_stateid
*stp
, struct nfsd4_open
*open
) {
1246 unsigned int access
, deny
;
1248 set_access(&access
, stp
->st_access_bmap
);
1249 set_deny(&deny
, stp
->st_deny_bmap
);
1250 if ((access
& open
->op_share_deny
) || (deny
& open
->op_share_access
))
1256 * Called to check deny when READ with all zero stateid or
1257 * WRITE with all zero or all one stateid
1260 nfs4_share_conflict(struct svc_fh
*current_fh
, unsigned int deny_type
)
1262 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
1263 struct nfs4_file
*fp
;
1264 struct nfs4_stateid
*stp
;
1267 dprintk("NFSD: nfs4_share_conflict\n");
1269 fp
= find_file(ino
);
1272 ret
= nfserr_locked
;
1273 /* Search for conflicting share reservations */
1274 list_for_each_entry(stp
, &fp
->fi_stateids
, st_perfile
) {
1275 if (test_bit(deny_type
, &stp
->st_deny_bmap
) ||
1276 test_bit(NFS4_SHARE_DENY_BOTH
, &stp
->st_deny_bmap
))
1286 nfs4_file_downgrade(struct file
*filp
, unsigned int share_access
)
1288 if (share_access
& NFS4_SHARE_ACCESS_WRITE
) {
1289 drop_file_write_access(filp
);
1290 filp
->f_mode
= (filp
->f_mode
| FMODE_READ
) & ~FMODE_WRITE
;
1295 * Recall a delegation
1298 do_recall(void *__dp
)
1300 struct nfs4_delegation
*dp
= __dp
;
1302 dp
->dl_file
->fi_had_conflict
= true;
1303 nfsd4_cb_recall(dp
);
1308 * Spawn a thread to perform a recall on the delegation represented
1309 * by the lease (file_lock)
1311 * Called from break_lease() with lock_kernel() held.
1312 * Note: we assume break_lease will only call this *once* for any given
1316 void nfsd_break_deleg_cb(struct file_lock
*fl
)
1318 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
1319 struct task_struct
*t
;
1321 dprintk("NFSD nfsd_break_deleg_cb: dp %p fl %p\n",dp
,fl
);
1325 /* We're assuming the state code never drops its reference
1326 * without first removing the lease. Since we're in this lease
1327 * callback (and since the lease code is serialized by the kernel
1328 * lock) we know the server hasn't removed the lease yet, we know
1329 * it's safe to take a reference: */
1330 atomic_inc(&dp
->dl_count
);
1331 atomic_inc(&dp
->dl_client
->cl_count
);
1333 spin_lock(&recall_lock
);
1334 list_add_tail(&dp
->dl_recall_lru
, &del_recall_lru
);
1335 spin_unlock(&recall_lock
);
1337 /* only place dl_time is set. protected by lock_kernel*/
1338 dp
->dl_time
= get_seconds();
1341 * We don't want the locks code to timeout the lease for us;
1342 * we'll remove it ourself if the delegation isn't returned
1345 fl
->fl_break_time
= 0;
1347 t
= kthread_run(do_recall
, dp
, "%s", "nfs4_cb_recall");
1349 struct nfs4_client
*clp
= dp
->dl_client
;
1351 printk(KERN_INFO
"NFSD: Callback thread failed for "
1352 "for client (clientid %08x/%08x)\n",
1353 clp
->cl_clientid
.cl_boot
, clp
->cl_clientid
.cl_id
);
1354 put_nfs4_client(dp
->dl_client
);
1355 nfs4_put_delegation(dp
);
1360 * The file_lock is being reapd.
1362 * Called by locks_free_lock() with lock_kernel() held.
1365 void nfsd_release_deleg_cb(struct file_lock
*fl
)
1367 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)fl
->fl_owner
;
1369 dprintk("NFSD nfsd_release_deleg_cb: fl %p dp %p dl_count %d\n", fl
,dp
, atomic_read(&dp
->dl_count
));
1371 if (!(fl
->fl_flags
& FL_LEASE
) || !dp
)
1373 dp
->dl_flock
= NULL
;
1377 * Set the delegation file_lock back pointer.
1379 * Called from setlease() with lock_kernel() held.
1382 void nfsd_copy_lock_deleg_cb(struct file_lock
*new, struct file_lock
*fl
)
1384 struct nfs4_delegation
*dp
= (struct nfs4_delegation
*)new->fl_owner
;
1386 dprintk("NFSD: nfsd_copy_lock_deleg_cb: new fl %p dp %p\n", new, dp
);
1393 * Called from setlease() with lock_kernel() held
1396 int nfsd_same_client_deleg_cb(struct file_lock
*onlist
, struct file_lock
*try)
1398 struct nfs4_delegation
*onlistd
=
1399 (struct nfs4_delegation
*)onlist
->fl_owner
;
1400 struct nfs4_delegation
*tryd
=
1401 (struct nfs4_delegation
*)try->fl_owner
;
1403 if (onlist
->fl_lmops
!= try->fl_lmops
)
1406 return onlistd
->dl_client
== tryd
->dl_client
;
1411 int nfsd_change_deleg_cb(struct file_lock
**onlist
, int arg
)
1414 return lease_modify(onlist
, arg
);
1419 static struct lock_manager_operations nfsd_lease_mng_ops
= {
1420 .fl_break
= nfsd_break_deleg_cb
,
1421 .fl_release_private
= nfsd_release_deleg_cb
,
1422 .fl_copy_lock
= nfsd_copy_lock_deleg_cb
,
1423 .fl_mylease
= nfsd_same_client_deleg_cb
,
1424 .fl_change
= nfsd_change_deleg_cb
,
1429 nfsd4_process_open1(struct nfsd4_open
*open
)
1431 clientid_t
*clientid
= &open
->op_clientid
;
1432 struct nfs4_client
*clp
= NULL
;
1433 unsigned int strhashval
;
1434 struct nfs4_stateowner
*sop
= NULL
;
1436 if (!check_name(open
->op_owner
))
1437 return nfserr_inval
;
1439 if (STALE_CLIENTID(&open
->op_clientid
))
1440 return nfserr_stale_clientid
;
1442 strhashval
= ownerstr_hashval(clientid
->cl_id
, open
->op_owner
);
1443 sop
= find_openstateowner_str(strhashval
, open
);
1444 open
->op_stateowner
= sop
;
1446 /* Make sure the client's lease hasn't expired. */
1447 clp
= find_confirmed_client(clientid
);
1449 return nfserr_expired
;
1452 if (!sop
->so_confirmed
) {
1453 /* Replace unconfirmed owners without checking for replay. */
1454 clp
= sop
->so_client
;
1455 release_openowner(sop
);
1456 open
->op_stateowner
= NULL
;
1459 if (open
->op_seqid
== sop
->so_seqid
- 1) {
1460 if (sop
->so_replay
.rp_buflen
)
1461 return nfserr_replay_me
;
1462 /* The original OPEN failed so spectacularly
1463 * that we don't even have replay data saved!
1464 * Therefore, we have no choice but to continue
1465 * processing this OPEN; presumably, we'll
1466 * fail again for the same reason.
1468 dprintk("nfsd4_process_open1: replay with no replay cache\n");
1471 if (open
->op_seqid
!= sop
->so_seqid
)
1472 return nfserr_bad_seqid
;
1474 if (open
->op_stateowner
== NULL
) {
1475 sop
= alloc_init_open_stateowner(strhashval
, clp
, open
);
1477 return nfserr_resource
;
1478 open
->op_stateowner
= sop
;
1480 list_del_init(&sop
->so_close_lru
);
1481 renew_client(sop
->so_client
);
1485 static inline __be32
1486 nfs4_check_delegmode(struct nfs4_delegation
*dp
, int flags
)
1488 if ((flags
& WR_STATE
) && (dp
->dl_type
== NFS4_OPEN_DELEGATE_READ
))
1489 return nfserr_openmode
;
1494 static struct nfs4_delegation
*
1495 find_delegation_file(struct nfs4_file
*fp
, stateid_t
*stid
)
1497 struct nfs4_delegation
*dp
;
1499 list_for_each_entry(dp
, &fp
->fi_delegations
, dl_perfile
) {
1500 if (dp
->dl_stateid
.si_stateownerid
== stid
->si_stateownerid
)
1507 nfs4_check_deleg(struct nfs4_file
*fp
, struct nfsd4_open
*open
,
1508 struct nfs4_delegation
**dp
)
1511 __be32 status
= nfserr_bad_stateid
;
1513 *dp
= find_delegation_file(fp
, &open
->op_delegate_stateid
);
1516 flags
= open
->op_share_access
== NFS4_SHARE_ACCESS_READ
?
1517 RD_STATE
: WR_STATE
;
1518 status
= nfs4_check_delegmode(*dp
, flags
);
1522 if (open
->op_claim_type
!= NFS4_OPEN_CLAIM_DELEGATE_CUR
)
1526 open
->op_stateowner
->so_confirmed
= 1;
1531 nfs4_check_open(struct nfs4_file
*fp
, struct nfsd4_open
*open
, struct nfs4_stateid
**stpp
)
1533 struct nfs4_stateid
*local
;
1534 __be32 status
= nfserr_share_denied
;
1535 struct nfs4_stateowner
*sop
= open
->op_stateowner
;
1537 list_for_each_entry(local
, &fp
->fi_stateids
, st_perfile
) {
1538 /* ignore lock owners */
1539 if (local
->st_stateowner
->so_is_open_owner
== 0)
1541 /* remember if we have seen this open owner */
1542 if (local
->st_stateowner
== sop
)
1544 /* check for conflicting share reservations */
1545 if (!test_share(local
, open
))
1553 static inline struct nfs4_stateid
*
1554 nfs4_alloc_stateid(void)
1556 return kmem_cache_alloc(stateid_slab
, GFP_KERNEL
);
1560 nfs4_new_open(struct svc_rqst
*rqstp
, struct nfs4_stateid
**stpp
,
1561 struct nfs4_delegation
*dp
,
1562 struct svc_fh
*cur_fh
, int flags
)
1564 struct nfs4_stateid
*stp
;
1566 stp
= nfs4_alloc_stateid();
1568 return nfserr_resource
;
1571 get_file(dp
->dl_vfs_file
);
1572 stp
->st_vfs_file
= dp
->dl_vfs_file
;
1575 status
= nfsd_open(rqstp
, cur_fh
, S_IFREG
, flags
,
1578 if (status
== nfserr_dropit
)
1579 status
= nfserr_jukebox
;
1580 kmem_cache_free(stateid_slab
, stp
);
1588 static inline __be32
1589 nfsd4_truncate(struct svc_rqst
*rqstp
, struct svc_fh
*fh
,
1590 struct nfsd4_open
*open
)
1592 struct iattr iattr
= {
1593 .ia_valid
= ATTR_SIZE
,
1596 if (!open
->op_truncate
)
1598 if (!(open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
))
1599 return nfserr_inval
;
1600 return nfsd_setattr(rqstp
, fh
, &iattr
, 0, (time_t)0);
1604 nfs4_upgrade_open(struct svc_rqst
*rqstp
, struct svc_fh
*cur_fh
, struct nfs4_stateid
*stp
, struct nfsd4_open
*open
)
1606 struct file
*filp
= stp
->st_vfs_file
;
1607 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
1608 unsigned int share_access
, new_writer
;
1611 set_access(&share_access
, stp
->st_access_bmap
);
1612 new_writer
= (~share_access
) & open
->op_share_access
1613 & NFS4_SHARE_ACCESS_WRITE
;
1616 int err
= get_write_access(inode
);
1618 return nfserrno(err
);
1619 err
= mnt_want_write(cur_fh
->fh_export
->ex_path
.mnt
);
1621 return nfserrno(err
);
1622 file_take_write(filp
);
1624 status
= nfsd4_truncate(rqstp
, cur_fh
, open
);
1627 put_write_access(inode
);
1630 /* remember the open */
1631 filp
->f_mode
|= open
->op_share_access
;
1632 __set_bit(open
->op_share_access
, &stp
->st_access_bmap
);
1633 __set_bit(open
->op_share_deny
, &stp
->st_deny_bmap
);
1640 nfs4_set_claim_prev(struct nfsd4_open
*open
)
1642 open
->op_stateowner
->so_confirmed
= 1;
1643 open
->op_stateowner
->so_client
->cl_firststate
= 1;
1647 * Attempt to hand out a delegation.
1650 nfs4_open_delegation(struct svc_fh
*fh
, struct nfsd4_open
*open
, struct nfs4_stateid
*stp
)
1652 struct nfs4_delegation
*dp
;
1653 struct nfs4_stateowner
*sop
= stp
->st_stateowner
;
1654 struct nfs4_callback
*cb
= &sop
->so_client
->cl_callback
;
1655 struct file_lock fl
, *flp
= &fl
;
1656 int status
, flag
= 0;
1658 flag
= NFS4_OPEN_DELEGATE_NONE
;
1659 open
->op_recall
= 0;
1660 switch (open
->op_claim_type
) {
1661 case NFS4_OPEN_CLAIM_PREVIOUS
:
1662 if (!atomic_read(&cb
->cb_set
))
1663 open
->op_recall
= 1;
1664 flag
= open
->op_delegate_type
;
1665 if (flag
== NFS4_OPEN_DELEGATE_NONE
)
1668 case NFS4_OPEN_CLAIM_NULL
:
1669 /* Let's not give out any delegations till everyone's
1670 * had the chance to reclaim theirs.... */
1671 if (locks_in_grace())
1673 if (!atomic_read(&cb
->cb_set
) || !sop
->so_confirmed
)
1675 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
1676 flag
= NFS4_OPEN_DELEGATE_WRITE
;
1678 flag
= NFS4_OPEN_DELEGATE_READ
;
1684 dp
= alloc_init_deleg(sop
->so_client
, stp
, fh
, flag
);
1686 flag
= NFS4_OPEN_DELEGATE_NONE
;
1689 locks_init_lock(&fl
);
1690 fl
.fl_lmops
= &nfsd_lease_mng_ops
;
1691 fl
.fl_flags
= FL_LEASE
;
1692 fl
.fl_type
= flag
== NFS4_OPEN_DELEGATE_READ
? F_RDLCK
: F_WRLCK
;
1693 fl
.fl_end
= OFFSET_MAX
;
1694 fl
.fl_owner
= (fl_owner_t
)dp
;
1695 fl
.fl_file
= stp
->st_vfs_file
;
1696 fl
.fl_pid
= current
->tgid
;
1698 /* vfs_setlease checks to see if delegation should be handed out.
1699 * the lock_manager callbacks fl_mylease and fl_change are used
1701 if ((status
= vfs_setlease(stp
->st_vfs_file
, fl
.fl_type
, &flp
))) {
1702 dprintk("NFSD: setlease failed [%d], no delegation\n", status
);
1703 unhash_delegation(dp
);
1704 flag
= NFS4_OPEN_DELEGATE_NONE
;
1708 memcpy(&open
->op_delegate_stateid
, &dp
->dl_stateid
, sizeof(dp
->dl_stateid
));
1710 dprintk("NFSD: delegation stateid=(%08x/%08x/%08x/%08x)\n\n",
1711 dp
->dl_stateid
.si_boot
,
1712 dp
->dl_stateid
.si_stateownerid
,
1713 dp
->dl_stateid
.si_fileid
,
1714 dp
->dl_stateid
.si_generation
);
1716 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
1717 && flag
== NFS4_OPEN_DELEGATE_NONE
1718 && open
->op_delegate_type
!= NFS4_OPEN_DELEGATE_NONE
)
1719 dprintk("NFSD: WARNING: refusing delegation reclaim\n");
1720 open
->op_delegate_type
= flag
;
1724 * called with nfs4_lock_state() held.
1727 nfsd4_process_open2(struct svc_rqst
*rqstp
, struct svc_fh
*current_fh
, struct nfsd4_open
*open
)
1729 struct nfs4_file
*fp
= NULL
;
1730 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
1731 struct nfs4_stateid
*stp
= NULL
;
1732 struct nfs4_delegation
*dp
= NULL
;
1735 status
= nfserr_inval
;
1736 if (!access_valid(open
->op_share_access
)
1737 || !deny_valid(open
->op_share_deny
))
1740 * Lookup file; if found, lookup stateid and check open request,
1741 * and check for delegations in the process of being recalled.
1742 * If not found, create the nfs4_file struct
1744 fp
= find_file(ino
);
1746 if ((status
= nfs4_check_open(fp
, open
, &stp
)))
1748 status
= nfs4_check_deleg(fp
, open
, &dp
);
1752 status
= nfserr_bad_stateid
;
1753 if (open
->op_claim_type
== NFS4_OPEN_CLAIM_DELEGATE_CUR
)
1755 status
= nfserr_resource
;
1756 fp
= alloc_init_file(ino
);
1762 * OPEN the file, or upgrade an existing OPEN.
1763 * If truncate fails, the OPEN fails.
1766 /* Stateid was found, this is an OPEN upgrade */
1767 status
= nfs4_upgrade_open(rqstp
, current_fh
, stp
, open
);
1770 update_stateid(&stp
->st_stateid
);
1772 /* Stateid was not found, this is a new OPEN */
1774 if (open
->op_share_access
& NFS4_SHARE_ACCESS_READ
)
1775 flags
|= NFSD_MAY_READ
;
1776 if (open
->op_share_access
& NFS4_SHARE_ACCESS_WRITE
)
1777 flags
|= NFSD_MAY_WRITE
;
1778 status
= nfs4_new_open(rqstp
, &stp
, dp
, current_fh
, flags
);
1781 init_stateid(stp
, fp
, open
);
1782 status
= nfsd4_truncate(rqstp
, current_fh
, open
);
1784 release_open_stateid(stp
);
1788 memcpy(&open
->op_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
1791 * Attempt to hand out a delegation. No error return, because the
1792 * OPEN succeeds even if we fail.
1794 nfs4_open_delegation(current_fh
, open
, stp
);
1798 dprintk("nfs4_process_open2: stateid=(%08x/%08x/%08x/%08x)\n",
1799 stp
->st_stateid
.si_boot
, stp
->st_stateid
.si_stateownerid
,
1800 stp
->st_stateid
.si_fileid
, stp
->st_stateid
.si_generation
);
1804 if (status
== 0 && open
->op_claim_type
== NFS4_OPEN_CLAIM_PREVIOUS
)
1805 nfs4_set_claim_prev(open
);
1807 * To finish the open response, we just need to set the rflags.
1809 open
->op_rflags
= NFS4_OPEN_RESULT_LOCKTYPE_POSIX
;
1810 if (!open
->op_stateowner
->so_confirmed
)
1811 open
->op_rflags
|= NFS4_OPEN_RESULT_CONFIRM
;
1817 nfsd4_renew(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
1820 struct nfs4_client
*clp
;
1824 dprintk("process_renew(%08x/%08x): starting\n",
1825 clid
->cl_boot
, clid
->cl_id
);
1826 status
= nfserr_stale_clientid
;
1827 if (STALE_CLIENTID(clid
))
1829 clp
= find_confirmed_client(clid
);
1830 status
= nfserr_expired
;
1832 /* We assume the client took too long to RENEW. */
1833 dprintk("nfsd4_renew: clientid not found!\n");
1837 status
= nfserr_cb_path_down
;
1838 if (!list_empty(&clp
->cl_delegations
)
1839 && !atomic_read(&clp
->cl_callback
.cb_set
))
1843 nfs4_unlock_state();
1847 struct lock_manager nfsd4_manager
= {
1851 nfsd4_end_grace(void)
1853 dprintk("NFSD: end of grace period\n");
1854 nfsd4_recdir_purge_old();
1855 locks_end_grace(&nfsd4_manager
);
1859 nfs4_laundromat(void)
1861 struct nfs4_client
*clp
;
1862 struct nfs4_stateowner
*sop
;
1863 struct nfs4_delegation
*dp
;
1864 struct list_head
*pos
, *next
, reaplist
;
1865 time_t cutoff
= get_seconds() - NFSD_LEASE_TIME
;
1866 time_t t
, clientid_val
= NFSD_LEASE_TIME
;
1867 time_t u
, test_val
= NFSD_LEASE_TIME
;
1871 dprintk("NFSD: laundromat service - starting\n");
1872 if (locks_in_grace())
1874 list_for_each_safe(pos
, next
, &client_lru
) {
1875 clp
= list_entry(pos
, struct nfs4_client
, cl_lru
);
1876 if (time_after((unsigned long)clp
->cl_time
, (unsigned long)cutoff
)) {
1877 t
= clp
->cl_time
- cutoff
;
1878 if (clientid_val
> t
)
1882 dprintk("NFSD: purging unused client (clientid %08x)\n",
1883 clp
->cl_clientid
.cl_id
);
1884 nfsd4_remove_clid_dir(clp
);
1887 INIT_LIST_HEAD(&reaplist
);
1888 spin_lock(&recall_lock
);
1889 list_for_each_safe(pos
, next
, &del_recall_lru
) {
1890 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
1891 if (time_after((unsigned long)dp
->dl_time
, (unsigned long)cutoff
)) {
1892 u
= dp
->dl_time
- cutoff
;
1897 dprintk("NFSD: purging unused delegation dp %p, fp %p\n",
1899 list_move(&dp
->dl_recall_lru
, &reaplist
);
1901 spin_unlock(&recall_lock
);
1902 list_for_each_safe(pos
, next
, &reaplist
) {
1903 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
1904 list_del_init(&dp
->dl_recall_lru
);
1905 unhash_delegation(dp
);
1907 test_val
= NFSD_LEASE_TIME
;
1908 list_for_each_safe(pos
, next
, &close_lru
) {
1909 sop
= list_entry(pos
, struct nfs4_stateowner
, so_close_lru
);
1910 if (time_after((unsigned long)sop
->so_time
, (unsigned long)cutoff
)) {
1911 u
= sop
->so_time
- cutoff
;
1916 dprintk("NFSD: purging unused open stateowner (so_id %d)\n",
1918 release_openowner(sop
);
1920 if (clientid_val
< NFSD_LAUNDROMAT_MINTIMEOUT
)
1921 clientid_val
= NFSD_LAUNDROMAT_MINTIMEOUT
;
1922 nfs4_unlock_state();
1923 return clientid_val
;
1926 static struct workqueue_struct
*laundry_wq
;
1927 static void laundromat_main(struct work_struct
*);
1928 static DECLARE_DELAYED_WORK(laundromat_work
, laundromat_main
);
1931 laundromat_main(struct work_struct
*not_used
)
1935 t
= nfs4_laundromat();
1936 dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t
);
1937 queue_delayed_work(laundry_wq
, &laundromat_work
, t
*HZ
);
1940 static struct nfs4_stateowner
*
1941 search_close_lru(u32 st_id
, int flags
)
1943 struct nfs4_stateowner
*local
= NULL
;
1945 if (flags
& CLOSE_STATE
) {
1946 list_for_each_entry(local
, &close_lru
, so_close_lru
) {
1947 if (local
->so_id
== st_id
)
1955 nfs4_check_fh(struct svc_fh
*fhp
, struct nfs4_stateid
*stp
)
1957 return fhp
->fh_dentry
->d_inode
!= stp
->st_vfs_file
->f_path
.dentry
->d_inode
;
1961 STALE_STATEID(stateid_t
*stateid
)
1963 if (stateid
->si_boot
== boot_time
)
1965 dprintk("NFSD: stale stateid (%08x/%08x/%08x/%08x)!\n",
1966 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
1967 stateid
->si_generation
);
1972 access_permit_read(unsigned long access_bmap
)
1974 return test_bit(NFS4_SHARE_ACCESS_READ
, &access_bmap
) ||
1975 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
) ||
1976 test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
);
1980 access_permit_write(unsigned long access_bmap
)
1982 return test_bit(NFS4_SHARE_ACCESS_WRITE
, &access_bmap
) ||
1983 test_bit(NFS4_SHARE_ACCESS_BOTH
, &access_bmap
);
1987 __be32
nfs4_check_openmode(struct nfs4_stateid
*stp
, int flags
)
1989 __be32 status
= nfserr_openmode
;
1991 if ((flags
& WR_STATE
) && (!access_permit_write(stp
->st_access_bmap
)))
1993 if ((flags
& RD_STATE
) && (!access_permit_read(stp
->st_access_bmap
)))
2000 static inline __be32
2001 check_special_stateids(svc_fh
*current_fh
, stateid_t
*stateid
, int flags
)
2003 /* Trying to call delegreturn with a special stateid? Yuch: */
2004 if (!(flags
& (RD_STATE
| WR_STATE
)))
2005 return nfserr_bad_stateid
;
2006 else if (ONE_STATEID(stateid
) && (flags
& RD_STATE
))
2008 else if (locks_in_grace()) {
2009 /* Answer in remaining cases depends on existance of
2010 * conflicting state; so we must wait out the grace period. */
2011 return nfserr_grace
;
2012 } else if (flags
& WR_STATE
)
2013 return nfs4_share_conflict(current_fh
,
2014 NFS4_SHARE_DENY_WRITE
);
2015 else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
2016 return nfs4_share_conflict(current_fh
,
2017 NFS4_SHARE_DENY_READ
);
2021 * Allow READ/WRITE during grace period on recovered state only for files
2022 * that are not able to provide mandatory locking.
2025 io_during_grace_disallowed(struct inode
*inode
, int flags
)
2027 return locks_in_grace() && (flags
& (RD_STATE
| WR_STATE
))
2028 && mandatory_lock(inode
);
2031 static int check_stateid_generation(stateid_t
*in
, stateid_t
*ref
)
2033 /* If the client sends us a stateid from the future, it's buggy: */
2034 if (in
->si_generation
> ref
->si_generation
)
2035 return nfserr_bad_stateid
;
2037 * The following, however, can happen. For example, if the
2038 * client sends an open and some IO at the same time, the open
2039 * may bump si_generation while the IO is still in flight.
2040 * Thanks to hard links and renames, the client never knows what
2041 * file an open will affect. So it could avoid that situation
2042 * only by serializing all opens and IO from the same open
2043 * owner. To recover from the old_stateid error, the client
2044 * will just have to retry the IO:
2046 if (in
->si_generation
< ref
->si_generation
)
2047 return nfserr_old_stateid
;
2051 static int is_delegation_stateid(stateid_t
*stateid
)
2053 return stateid
->si_fileid
== 0;
2057 * Checks for stateid operations
2060 nfs4_preprocess_stateid_op(struct svc_fh
*current_fh
, stateid_t
*stateid
, int flags
, struct file
**filpp
)
2062 struct nfs4_stateid
*stp
= NULL
;
2063 struct nfs4_delegation
*dp
= NULL
;
2064 struct inode
*ino
= current_fh
->fh_dentry
->d_inode
;
2070 if (io_during_grace_disallowed(ino
, flags
))
2071 return nfserr_grace
;
2073 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
))
2074 return check_special_stateids(current_fh
, stateid
, flags
);
2076 status
= nfserr_stale_stateid
;
2077 if (STALE_STATEID(stateid
))
2080 status
= nfserr_bad_stateid
;
2081 if (is_delegation_stateid(stateid
)) {
2082 dp
= find_delegation_stateid(ino
, stateid
);
2085 status
= check_stateid_generation(stateid
, &dp
->dl_stateid
);
2088 status
= nfs4_check_delegmode(dp
, flags
);
2091 renew_client(dp
->dl_client
);
2092 if (flags
& DELEG_RET
)
2093 unhash_delegation(dp
);
2095 *filpp
= dp
->dl_vfs_file
;
2096 } else { /* open or lock stateid */
2097 stp
= find_stateid(stateid
, flags
);
2100 if ((flags
& CHECK_FH
) && nfs4_check_fh(current_fh
, stp
))
2102 if (!stp
->st_stateowner
->so_confirmed
)
2104 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2107 status
= nfs4_check_openmode(stp
, flags
);
2110 renew_client(stp
->st_stateowner
->so_client
);
2112 *filpp
= stp
->st_vfs_file
;
2122 return (type
== NFS4_READW_LT
|| type
== NFS4_READ_LT
) ?
2123 RD_STATE
: WR_STATE
;
2127 * Checks for sequence id mutating operations.
2130 nfs4_preprocess_seqid_op(struct svc_fh
*current_fh
, u32 seqid
, stateid_t
*stateid
, int flags
, struct nfs4_stateowner
**sopp
, struct nfs4_stateid
**stpp
, struct nfsd4_lock
*lock
)
2132 struct nfs4_stateid
*stp
;
2133 struct nfs4_stateowner
*sop
;
2136 dprintk("NFSD: preprocess_seqid_op: seqid=%d "
2137 "stateid = (%08x/%08x/%08x/%08x)\n", seqid
,
2138 stateid
->si_boot
, stateid
->si_stateownerid
, stateid
->si_fileid
,
2139 stateid
->si_generation
);
2144 if (ZERO_STATEID(stateid
) || ONE_STATEID(stateid
)) {
2145 dprintk("NFSD: preprocess_seqid_op: magic stateid!\n");
2146 return nfserr_bad_stateid
;
2149 if (STALE_STATEID(stateid
))
2150 return nfserr_stale_stateid
;
2152 * We return BAD_STATEID if filehandle doesn't match stateid,
2153 * the confirmed flag is incorrecly set, or the generation
2154 * number is incorrect.
2156 stp
= find_stateid(stateid
, flags
);
2159 * Also, we should make sure this isn't just the result of
2162 sop
= search_close_lru(stateid
->si_stateownerid
, flags
);
2164 return nfserr_bad_stateid
;
2170 *sopp
= sop
= stp
->st_stateowner
;
2173 clientid_t
*lockclid
= &lock
->v
.new.clientid
;
2174 struct nfs4_client
*clp
= sop
->so_client
;
2178 lkflg
= setlkflg(lock
->lk_type
);
2180 if (lock
->lk_is_new
) {
2181 if (!sop
->so_is_open_owner
)
2182 return nfserr_bad_stateid
;
2183 if (!same_clid(&clp
->cl_clientid
, lockclid
))
2184 return nfserr_bad_stateid
;
2185 /* stp is the open stateid */
2186 status
= nfs4_check_openmode(stp
, lkflg
);
2190 /* stp is the lock stateid */
2191 status
= nfs4_check_openmode(stp
->st_openstp
, lkflg
);
2197 if (nfs4_check_fh(current_fh
, stp
)) {
2198 dprintk("NFSD: preprocess_seqid_op: fh-stateid mismatch!\n");
2199 return nfserr_bad_stateid
;
2203 * We now validate the seqid and stateid generation numbers.
2204 * For the moment, we ignore the possibility of
2205 * generation number wraparound.
2207 if (seqid
!= sop
->so_seqid
)
2210 if (sop
->so_confirmed
&& flags
& CONFIRM
) {
2211 dprintk("NFSD: preprocess_seqid_op: expected"
2212 " unconfirmed stateowner!\n");
2213 return nfserr_bad_stateid
;
2215 if (!sop
->so_confirmed
&& !(flags
& CONFIRM
)) {
2216 dprintk("NFSD: preprocess_seqid_op: stateowner not"
2217 " confirmed yet!\n");
2218 return nfserr_bad_stateid
;
2220 status
= check_stateid_generation(stateid
, &stp
->st_stateid
);
2223 renew_client(sop
->so_client
);
2227 if (seqid
== sop
->so_seqid
- 1) {
2228 dprintk("NFSD: preprocess_seqid_op: retransmission?\n");
2229 /* indicate replay to calling function */
2230 return nfserr_replay_me
;
2232 dprintk("NFSD: preprocess_seqid_op: bad seqid (expected %d, got %d)\n",
2233 sop
->so_seqid
, seqid
);
2235 return nfserr_bad_seqid
;
2239 nfsd4_open_confirm(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2240 struct nfsd4_open_confirm
*oc
)
2243 struct nfs4_stateowner
*sop
;
2244 struct nfs4_stateid
*stp
;
2246 dprintk("NFSD: nfsd4_open_confirm on file %.*s\n",
2247 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2248 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2250 status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0);
2256 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2257 oc
->oc_seqid
, &oc
->oc_req_stateid
,
2258 CONFIRM
| OPEN_STATE
,
2259 &oc
->oc_stateowner
, &stp
, NULL
)))
2262 sop
= oc
->oc_stateowner
;
2263 sop
->so_confirmed
= 1;
2264 update_stateid(&stp
->st_stateid
);
2265 memcpy(&oc
->oc_resp_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2266 dprintk("NFSD: nfsd4_open_confirm: success, seqid=%d "
2267 "stateid=(%08x/%08x/%08x/%08x)\n", oc
->oc_seqid
,
2268 stp
->st_stateid
.si_boot
,
2269 stp
->st_stateid
.si_stateownerid
,
2270 stp
->st_stateid
.si_fileid
,
2271 stp
->st_stateid
.si_generation
);
2273 nfsd4_create_clid_dir(sop
->so_client
);
2275 if (oc
->oc_stateowner
) {
2276 nfs4_get_stateowner(oc
->oc_stateowner
);
2277 cstate
->replay_owner
= oc
->oc_stateowner
;
2279 nfs4_unlock_state();
2285 * unset all bits in union bitmap (bmap) that
2286 * do not exist in share (from successful OPEN_DOWNGRADE)
2289 reset_union_bmap_access(unsigned long access
, unsigned long *bmap
)
2292 for (i
= 1; i
< 4; i
++) {
2293 if ((i
& access
) != i
)
2294 __clear_bit(i
, bmap
);
2299 reset_union_bmap_deny(unsigned long deny
, unsigned long *bmap
)
2302 for (i
= 0; i
< 4; i
++) {
2303 if ((i
& deny
) != i
)
2304 __clear_bit(i
, bmap
);
2309 nfsd4_open_downgrade(struct svc_rqst
*rqstp
,
2310 struct nfsd4_compound_state
*cstate
,
2311 struct nfsd4_open_downgrade
*od
)
2314 struct nfs4_stateid
*stp
;
2315 unsigned int share_access
;
2317 dprintk("NFSD: nfsd4_open_downgrade on file %.*s\n",
2318 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2319 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2321 if (!access_valid(od
->od_share_access
)
2322 || !deny_valid(od
->od_share_deny
))
2323 return nfserr_inval
;
2326 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2330 &od
->od_stateowner
, &stp
, NULL
)))
2333 status
= nfserr_inval
;
2334 if (!test_bit(od
->od_share_access
, &stp
->st_access_bmap
)) {
2335 dprintk("NFSD:access not a subset current bitmap: 0x%lx, input access=%08x\n",
2336 stp
->st_access_bmap
, od
->od_share_access
);
2339 if (!test_bit(od
->od_share_deny
, &stp
->st_deny_bmap
)) {
2340 dprintk("NFSD:deny not a subset current bitmap: 0x%lx, input deny=%08x\n",
2341 stp
->st_deny_bmap
, od
->od_share_deny
);
2344 set_access(&share_access
, stp
->st_access_bmap
);
2345 nfs4_file_downgrade(stp
->st_vfs_file
,
2346 share_access
& ~od
->od_share_access
);
2348 reset_union_bmap_access(od
->od_share_access
, &stp
->st_access_bmap
);
2349 reset_union_bmap_deny(od
->od_share_deny
, &stp
->st_deny_bmap
);
2351 update_stateid(&stp
->st_stateid
);
2352 memcpy(&od
->od_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2355 if (od
->od_stateowner
) {
2356 nfs4_get_stateowner(od
->od_stateowner
);
2357 cstate
->replay_owner
= od
->od_stateowner
;
2359 nfs4_unlock_state();
2364 * nfs4_unlock_state() called after encode
2367 nfsd4_close(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2368 struct nfsd4_close
*close
)
2371 struct nfs4_stateid
*stp
;
2373 dprintk("NFSD: nfsd4_close on file %.*s\n",
2374 (int)cstate
->current_fh
.fh_dentry
->d_name
.len
,
2375 cstate
->current_fh
.fh_dentry
->d_name
.name
);
2378 /* check close_lru for replay */
2379 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2382 OPEN_STATE
| CLOSE_STATE
,
2383 &close
->cl_stateowner
, &stp
, NULL
)))
2386 update_stateid(&stp
->st_stateid
);
2387 memcpy(&close
->cl_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2389 /* release_stateid() calls nfsd_close() if needed */
2390 release_open_stateid(stp
);
2392 /* place unused nfs4_stateowners on so_close_lru list to be
2393 * released by the laundromat service after the lease period
2394 * to enable us to handle CLOSE replay
2396 if (list_empty(&close
->cl_stateowner
->so_stateids
))
2397 move_to_close_lru(close
->cl_stateowner
);
2399 if (close
->cl_stateowner
) {
2400 nfs4_get_stateowner(close
->cl_stateowner
);
2401 cstate
->replay_owner
= close
->cl_stateowner
;
2403 nfs4_unlock_state();
2408 nfsd4_delegreturn(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2409 struct nfsd4_delegreturn
*dr
)
2413 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0)))
2417 status
= nfs4_preprocess_stateid_op(&cstate
->current_fh
,
2418 &dr
->dr_stateid
, DELEG_RET
, NULL
);
2419 nfs4_unlock_state();
2426 * Lock owner state (byte-range locks)
2428 #define LOFF_OVERFLOW(start, len) ((u64)(len) > ~(u64)(start))
2429 #define LOCK_HASH_BITS 8
2430 #define LOCK_HASH_SIZE (1 << LOCK_HASH_BITS)
2431 #define LOCK_HASH_MASK (LOCK_HASH_SIZE - 1)
2434 end_offset(u64 start
, u64 len
)
2439 return end
>= start
? end
: NFS4_MAX_UINT64
;
2442 /* last octet in a range */
2444 last_byte_offset(u64 start
, u64 len
)
2450 return end
> start
? end
- 1: NFS4_MAX_UINT64
;
2453 #define lockownerid_hashval(id) \
2454 ((id) & LOCK_HASH_MASK)
2456 static inline unsigned int
2457 lock_ownerstr_hashval(struct inode
*inode
, u32 cl_id
,
2458 struct xdr_netobj
*ownername
)
2460 return (file_hashval(inode
) + cl_id
2461 + opaque_hashval(ownername
->data
, ownername
->len
))
2465 static struct list_head lock_ownerid_hashtbl
[LOCK_HASH_SIZE
];
2466 static struct list_head lock_ownerstr_hashtbl
[LOCK_HASH_SIZE
];
2467 static struct list_head lockstateid_hashtbl
[STATEID_HASH_SIZE
];
2469 static struct nfs4_stateid
*
2470 find_stateid(stateid_t
*stid
, int flags
)
2472 struct nfs4_stateid
*local
;
2473 u32 st_id
= stid
->si_stateownerid
;
2474 u32 f_id
= stid
->si_fileid
;
2475 unsigned int hashval
;
2477 dprintk("NFSD: find_stateid flags 0x%x\n",flags
);
2478 if (flags
& (LOCK_STATE
| RD_STATE
| WR_STATE
)) {
2479 hashval
= stateid_hashval(st_id
, f_id
);
2480 list_for_each_entry(local
, &lockstateid_hashtbl
[hashval
], st_hash
) {
2481 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
2482 (local
->st_stateid
.si_fileid
== f_id
))
2487 if (flags
& (OPEN_STATE
| RD_STATE
| WR_STATE
)) {
2488 hashval
= stateid_hashval(st_id
, f_id
);
2489 list_for_each_entry(local
, &stateid_hashtbl
[hashval
], st_hash
) {
2490 if ((local
->st_stateid
.si_stateownerid
== st_id
) &&
2491 (local
->st_stateid
.si_fileid
== f_id
))
2498 static struct nfs4_delegation
*
2499 find_delegation_stateid(struct inode
*ino
, stateid_t
*stid
)
2501 struct nfs4_file
*fp
;
2502 struct nfs4_delegation
*dl
;
2504 dprintk("NFSD:find_delegation_stateid stateid=(%08x/%08x/%08x/%08x)\n",
2505 stid
->si_boot
, stid
->si_stateownerid
,
2506 stid
->si_fileid
, stid
->si_generation
);
2508 fp
= find_file(ino
);
2511 dl
= find_delegation_file(fp
, stid
);
2517 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
2518 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
2519 * byte, because of sign extension problems. Since NFSv4 calls for 64-bit
2520 * locking, this prevents us from being completely protocol-compliant. The
2521 * real solution to this problem is to start using unsigned file offsets in
2522 * the VFS, but this is a very deep change!
2525 nfs4_transform_lock_offset(struct file_lock
*lock
)
2527 if (lock
->fl_start
< 0)
2528 lock
->fl_start
= OFFSET_MAX
;
2529 if (lock
->fl_end
< 0)
2530 lock
->fl_end
= OFFSET_MAX
;
2533 /* Hack!: For now, we're defining this just so we can use a pointer to it
2534 * as a unique cookie to identify our (NFSv4's) posix locks. */
2535 static struct lock_manager_operations nfsd_posix_mng_ops
= {
2539 nfs4_set_lock_denied(struct file_lock
*fl
, struct nfsd4_lock_denied
*deny
)
2541 struct nfs4_stateowner
*sop
;
2544 if (fl
->fl_lmops
== &nfsd_posix_mng_ops
) {
2545 sop
= (struct nfs4_stateowner
*) fl
->fl_owner
;
2546 hval
= lockownerid_hashval(sop
->so_id
);
2547 kref_get(&sop
->so_ref
);
2549 deny
->ld_clientid
= sop
->so_client
->cl_clientid
;
2551 deny
->ld_sop
= NULL
;
2552 deny
->ld_clientid
.cl_boot
= 0;
2553 deny
->ld_clientid
.cl_id
= 0;
2555 deny
->ld_start
= fl
->fl_start
;
2556 deny
->ld_length
= NFS4_MAX_UINT64
;
2557 if (fl
->fl_end
!= NFS4_MAX_UINT64
)
2558 deny
->ld_length
= fl
->fl_end
- fl
->fl_start
+ 1;
2559 deny
->ld_type
= NFS4_READ_LT
;
2560 if (fl
->fl_type
!= F_RDLCK
)
2561 deny
->ld_type
= NFS4_WRITE_LT
;
2564 static struct nfs4_stateowner
*
2565 find_lockstateowner_str(struct inode
*inode
, clientid_t
*clid
,
2566 struct xdr_netobj
*owner
)
2568 unsigned int hashval
= lock_ownerstr_hashval(inode
, clid
->cl_id
, owner
);
2569 struct nfs4_stateowner
*op
;
2571 list_for_each_entry(op
, &lock_ownerstr_hashtbl
[hashval
], so_strhash
) {
2572 if (same_owner_str(op
, owner
, clid
))
2579 * Alloc a lock owner structure.
2580 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
2583 * strhashval = lock_ownerstr_hashval
2586 static struct nfs4_stateowner
*
2587 alloc_init_lock_stateowner(unsigned int strhashval
, struct nfs4_client
*clp
, struct nfs4_stateid
*open_stp
, struct nfsd4_lock
*lock
) {
2588 struct nfs4_stateowner
*sop
;
2589 struct nfs4_replay
*rp
;
2590 unsigned int idhashval
;
2592 if (!(sop
= alloc_stateowner(&lock
->lk_new_owner
)))
2594 idhashval
= lockownerid_hashval(current_ownerid
);
2595 INIT_LIST_HEAD(&sop
->so_idhash
);
2596 INIT_LIST_HEAD(&sop
->so_strhash
);
2597 INIT_LIST_HEAD(&sop
->so_perclient
);
2598 INIT_LIST_HEAD(&sop
->so_stateids
);
2599 INIT_LIST_HEAD(&sop
->so_perstateid
);
2600 INIT_LIST_HEAD(&sop
->so_close_lru
); /* not used */
2602 list_add(&sop
->so_idhash
, &lock_ownerid_hashtbl
[idhashval
]);
2603 list_add(&sop
->so_strhash
, &lock_ownerstr_hashtbl
[strhashval
]);
2604 list_add(&sop
->so_perstateid
, &open_stp
->st_lockowners
);
2605 sop
->so_is_open_owner
= 0;
2606 sop
->so_id
= current_ownerid
++;
2607 sop
->so_client
= clp
;
2608 /* It is the openowner seqid that will be incremented in encode in the
2609 * case of new lockowners; so increment the lock seqid manually: */
2610 sop
->so_seqid
= lock
->lk_new_lock_seqid
+ 1;
2611 sop
->so_confirmed
= 1;
2612 rp
= &sop
->so_replay
;
2613 rp
->rp_status
= nfserr_serverfault
;
2615 rp
->rp_buf
= rp
->rp_ibuf
;
2619 static struct nfs4_stateid
*
2620 alloc_init_lock_stateid(struct nfs4_stateowner
*sop
, struct nfs4_file
*fp
, struct nfs4_stateid
*open_stp
)
2622 struct nfs4_stateid
*stp
;
2623 unsigned int hashval
= stateid_hashval(sop
->so_id
, fp
->fi_id
);
2625 stp
= nfs4_alloc_stateid();
2628 INIT_LIST_HEAD(&stp
->st_hash
);
2629 INIT_LIST_HEAD(&stp
->st_perfile
);
2630 INIT_LIST_HEAD(&stp
->st_perstateowner
);
2631 INIT_LIST_HEAD(&stp
->st_lockowners
); /* not used */
2632 list_add(&stp
->st_hash
, &lockstateid_hashtbl
[hashval
]);
2633 list_add(&stp
->st_perfile
, &fp
->fi_stateids
);
2634 list_add(&stp
->st_perstateowner
, &sop
->so_stateids
);
2635 stp
->st_stateowner
= sop
;
2638 stp
->st_stateid
.si_boot
= boot_time
;
2639 stp
->st_stateid
.si_stateownerid
= sop
->so_id
;
2640 stp
->st_stateid
.si_fileid
= fp
->fi_id
;
2641 stp
->st_stateid
.si_generation
= 0;
2642 stp
->st_vfs_file
= open_stp
->st_vfs_file
; /* FIXME refcount?? */
2643 stp
->st_access_bmap
= open_stp
->st_access_bmap
;
2644 stp
->st_deny_bmap
= open_stp
->st_deny_bmap
;
2645 stp
->st_openstp
= open_stp
;
2652 check_lock_length(u64 offset
, u64 length
)
2654 return ((length
== 0) || ((length
!= NFS4_MAX_UINT64
) &&
2655 LOFF_OVERFLOW(offset
, length
)));
2662 nfsd4_lock(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2663 struct nfsd4_lock
*lock
)
2665 struct nfs4_stateowner
*open_sop
= NULL
;
2666 struct nfs4_stateowner
*lock_sop
= NULL
;
2667 struct nfs4_stateid
*lock_stp
;
2669 struct file_lock file_lock
;
2670 struct file_lock conflock
;
2672 unsigned int strhashval
;
2676 dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
2677 (long long) lock
->lk_offset
,
2678 (long long) lock
->lk_length
);
2680 if (check_lock_length(lock
->lk_offset
, lock
->lk_length
))
2681 return nfserr_inval
;
2683 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
,
2684 S_IFREG
, NFSD_MAY_LOCK
))) {
2685 dprintk("NFSD: nfsd4_lock: permission denied!\n");
2691 if (lock
->lk_is_new
) {
2693 * Client indicates that this is a new lockowner.
2694 * Use open owner and open stateid to create lock owner and
2697 struct nfs4_stateid
*open_stp
= NULL
;
2698 struct nfs4_file
*fp
;
2700 status
= nfserr_stale_clientid
;
2701 if (STALE_CLIENTID(&lock
->lk_new_clientid
))
2704 /* validate and update open stateid and open seqid */
2705 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2706 lock
->lk_new_open_seqid
,
2707 &lock
->lk_new_open_stateid
,
2709 &lock
->lk_replay_owner
, &open_stp
,
2713 open_sop
= lock
->lk_replay_owner
;
2714 /* create lockowner and lock stateid */
2715 fp
= open_stp
->st_file
;
2716 strhashval
= lock_ownerstr_hashval(fp
->fi_inode
,
2717 open_sop
->so_client
->cl_clientid
.cl_id
,
2718 &lock
->v
.new.owner
);
2719 /* XXX: Do we need to check for duplicate stateowners on
2720 * the same file, or should they just be allowed (and
2721 * create new stateids)? */
2722 status
= nfserr_resource
;
2723 lock_sop
= alloc_init_lock_stateowner(strhashval
,
2724 open_sop
->so_client
, open_stp
, lock
);
2725 if (lock_sop
== NULL
)
2727 lock_stp
= alloc_init_lock_stateid(lock_sop
, fp
, open_stp
);
2728 if (lock_stp
== NULL
)
2731 /* lock (lock owner + lock stateid) already exists */
2732 status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2733 lock
->lk_old_lock_seqid
,
2734 &lock
->lk_old_lock_stateid
,
2736 &lock
->lk_replay_owner
, &lock_stp
, lock
);
2739 lock_sop
= lock
->lk_replay_owner
;
2741 /* lock->lk_replay_owner and lock_stp have been created or found */
2742 filp
= lock_stp
->st_vfs_file
;
2744 status
= nfserr_grace
;
2745 if (locks_in_grace() && !lock
->lk_reclaim
)
2747 status
= nfserr_no_grace
;
2748 if (!locks_in_grace() && lock
->lk_reclaim
)
2751 locks_init_lock(&file_lock
);
2752 switch (lock
->lk_type
) {
2755 file_lock
.fl_type
= F_RDLCK
;
2759 case NFS4_WRITEW_LT
:
2760 file_lock
.fl_type
= F_WRLCK
;
2764 status
= nfserr_inval
;
2767 file_lock
.fl_owner
= (fl_owner_t
)lock_sop
;
2768 file_lock
.fl_pid
= current
->tgid
;
2769 file_lock
.fl_file
= filp
;
2770 file_lock
.fl_flags
= FL_POSIX
;
2771 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
2773 file_lock
.fl_start
= lock
->lk_offset
;
2774 file_lock
.fl_end
= last_byte_offset(lock
->lk_offset
, lock
->lk_length
);
2775 nfs4_transform_lock_offset(&file_lock
);
2778 * Try to lock the file in the VFS.
2779 * Note: locks.c uses the BKL to protect the inode's lock list.
2782 err
= vfs_lock_file(filp
, cmd
, &file_lock
, &conflock
);
2784 case 0: /* success! */
2785 update_stateid(&lock_stp
->st_stateid
);
2786 memcpy(&lock
->lk_resp_stateid
, &lock_stp
->st_stateid
,
2790 case (EAGAIN
): /* conflock holds conflicting lock */
2791 status
= nfserr_denied
;
2792 dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
2793 nfs4_set_lock_denied(&conflock
, &lock
->lk_denied
);
2796 status
= nfserr_deadlock
;
2799 dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err
);
2800 status
= nfserr_resource
;
2804 if (status
&& lock
->lk_is_new
&& lock_sop
)
2805 release_lockowner(lock_sop
);
2806 if (lock
->lk_replay_owner
) {
2807 nfs4_get_stateowner(lock
->lk_replay_owner
);
2808 cstate
->replay_owner
= lock
->lk_replay_owner
;
2810 nfs4_unlock_state();
2815 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
2816 * so we do a temporary open here just to get an open file to pass to
2817 * vfs_test_lock. (Arguably perhaps test_lock should be done with an
2820 static int nfsd_test_lock(struct svc_rqst
*rqstp
, struct svc_fh
*fhp
, struct file_lock
*lock
)
2825 err
= nfsd_open(rqstp
, fhp
, S_IFREG
, NFSD_MAY_READ
, &file
);
2828 err
= vfs_test_lock(file
, lock
);
2837 nfsd4_lockt(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2838 struct nfsd4_lockt
*lockt
)
2840 struct inode
*inode
;
2841 struct file_lock file_lock
;
2845 if (locks_in_grace())
2846 return nfserr_grace
;
2848 if (check_lock_length(lockt
->lt_offset
, lockt
->lt_length
))
2849 return nfserr_inval
;
2851 lockt
->lt_stateowner
= NULL
;
2854 status
= nfserr_stale_clientid
;
2855 if (STALE_CLIENTID(&lockt
->lt_clientid
))
2858 if ((status
= fh_verify(rqstp
, &cstate
->current_fh
, S_IFREG
, 0))) {
2859 dprintk("NFSD: nfsd4_lockt: fh_verify() failed!\n");
2860 if (status
== nfserr_symlink
)
2861 status
= nfserr_inval
;
2865 inode
= cstate
->current_fh
.fh_dentry
->d_inode
;
2866 locks_init_lock(&file_lock
);
2867 switch (lockt
->lt_type
) {
2870 file_lock
.fl_type
= F_RDLCK
;
2873 case NFS4_WRITEW_LT
:
2874 file_lock
.fl_type
= F_WRLCK
;
2877 dprintk("NFSD: nfs4_lockt: bad lock type!\n");
2878 status
= nfserr_inval
;
2882 lockt
->lt_stateowner
= find_lockstateowner_str(inode
,
2883 &lockt
->lt_clientid
, &lockt
->lt_owner
);
2884 if (lockt
->lt_stateowner
)
2885 file_lock
.fl_owner
= (fl_owner_t
)lockt
->lt_stateowner
;
2886 file_lock
.fl_pid
= current
->tgid
;
2887 file_lock
.fl_flags
= FL_POSIX
;
2889 file_lock
.fl_start
= lockt
->lt_offset
;
2890 file_lock
.fl_end
= last_byte_offset(lockt
->lt_offset
, lockt
->lt_length
);
2892 nfs4_transform_lock_offset(&file_lock
);
2895 error
= nfsd_test_lock(rqstp
, &cstate
->current_fh
, &file_lock
);
2897 status
= nfserrno(error
);
2900 if (file_lock
.fl_type
!= F_UNLCK
) {
2901 status
= nfserr_denied
;
2902 nfs4_set_lock_denied(&file_lock
, &lockt
->lt_denied
);
2905 nfs4_unlock_state();
2910 nfsd4_locku(struct svc_rqst
*rqstp
, struct nfsd4_compound_state
*cstate
,
2911 struct nfsd4_locku
*locku
)
2913 struct nfs4_stateid
*stp
;
2914 struct file
*filp
= NULL
;
2915 struct file_lock file_lock
;
2919 dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
2920 (long long) locku
->lu_offset
,
2921 (long long) locku
->lu_length
);
2923 if (check_lock_length(locku
->lu_offset
, locku
->lu_length
))
2924 return nfserr_inval
;
2928 if ((status
= nfs4_preprocess_seqid_op(&cstate
->current_fh
,
2932 &locku
->lu_stateowner
, &stp
, NULL
)))
2935 filp
= stp
->st_vfs_file
;
2937 locks_init_lock(&file_lock
);
2938 file_lock
.fl_type
= F_UNLCK
;
2939 file_lock
.fl_owner
= (fl_owner_t
) locku
->lu_stateowner
;
2940 file_lock
.fl_pid
= current
->tgid
;
2941 file_lock
.fl_file
= filp
;
2942 file_lock
.fl_flags
= FL_POSIX
;
2943 file_lock
.fl_lmops
= &nfsd_posix_mng_ops
;
2944 file_lock
.fl_start
= locku
->lu_offset
;
2946 file_lock
.fl_end
= last_byte_offset(locku
->lu_offset
, locku
->lu_length
);
2947 nfs4_transform_lock_offset(&file_lock
);
2950 * Try to unlock the file in the VFS.
2952 err
= vfs_lock_file(filp
, F_SETLK
, &file_lock
, NULL
);
2954 dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
2958 * OK, unlock succeeded; the only thing left to do is update the stateid.
2960 update_stateid(&stp
->st_stateid
);
2961 memcpy(&locku
->lu_stateid
, &stp
->st_stateid
, sizeof(stateid_t
));
2964 if (locku
->lu_stateowner
) {
2965 nfs4_get_stateowner(locku
->lu_stateowner
);
2966 cstate
->replay_owner
= locku
->lu_stateowner
;
2968 nfs4_unlock_state();
2972 status
= nfserrno(err
);
2978 * 1: locks held by lockowner
2979 * 0: no locks held by lockowner
2982 check_for_locks(struct file
*filp
, struct nfs4_stateowner
*lowner
)
2984 struct file_lock
**flpp
;
2985 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
2989 for (flpp
= &inode
->i_flock
; *flpp
!= NULL
; flpp
= &(*flpp
)->fl_next
) {
2990 if ((*flpp
)->fl_owner
== (fl_owner_t
)lowner
) {
3001 nfsd4_release_lockowner(struct svc_rqst
*rqstp
,
3002 struct nfsd4_compound_state
*cstate
,
3003 struct nfsd4_release_lockowner
*rlockowner
)
3005 clientid_t
*clid
= &rlockowner
->rl_clientid
;
3006 struct nfs4_stateowner
*sop
;
3007 struct nfs4_stateid
*stp
;
3008 struct xdr_netobj
*owner
= &rlockowner
->rl_owner
;
3009 struct list_head matches
;
3013 dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
3014 clid
->cl_boot
, clid
->cl_id
);
3016 /* XXX check for lease expiration */
3018 status
= nfserr_stale_clientid
;
3019 if (STALE_CLIENTID(clid
))
3024 status
= nfserr_locks_held
;
3025 /* XXX: we're doing a linear search through all the lockowners.
3026 * Yipes! For now we'll just hope clients aren't really using
3027 * release_lockowner much, but eventually we have to fix these
3028 * data structures. */
3029 INIT_LIST_HEAD(&matches
);
3030 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3031 list_for_each_entry(sop
, &lock_ownerid_hashtbl
[i
], so_idhash
) {
3032 if (!same_owner_str(sop
, owner
, clid
))
3034 list_for_each_entry(stp
, &sop
->so_stateids
,
3036 if (check_for_locks(stp
->st_vfs_file
, sop
))
3038 /* Note: so_perclient unused for lockowners,
3039 * so it's OK to fool with here. */
3040 list_add(&sop
->so_perclient
, &matches
);
3044 /* Clients probably won't expect us to return with some (but not all)
3045 * of the lockowner state released; so don't release any until all
3046 * have been checked. */
3048 while (!list_empty(&matches
)) {
3049 sop
= list_entry(matches
.next
, struct nfs4_stateowner
,
3051 /* unhash_stateowner deletes so_perclient only
3052 * for openowners. */
3053 list_del(&sop
->so_perclient
);
3054 release_lockowner(sop
);
3057 nfs4_unlock_state();
3061 static inline struct nfs4_client_reclaim
*
3064 return kmalloc(sizeof(struct nfs4_client_reclaim
), GFP_KERNEL
);
3068 nfs4_has_reclaimed_state(const char *name
)
3070 unsigned int strhashval
= clientstr_hashval(name
);
3071 struct nfs4_client
*clp
;
3073 clp
= find_confirmed_client_by_str(name
, strhashval
);
3078 * failure => all reset bets are off, nfserr_no_grace...
3081 nfs4_client_to_reclaim(const char *name
)
3083 unsigned int strhashval
;
3084 struct nfs4_client_reclaim
*crp
= NULL
;
3086 dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN
, name
);
3087 crp
= alloc_reclaim();
3090 strhashval
= clientstr_hashval(name
);
3091 INIT_LIST_HEAD(&crp
->cr_strhash
);
3092 list_add(&crp
->cr_strhash
, &reclaim_str_hashtbl
[strhashval
]);
3093 memcpy(crp
->cr_recdir
, name
, HEXDIR_LEN
);
3094 reclaim_str_hashtbl_size
++;
3099 nfs4_release_reclaim(void)
3101 struct nfs4_client_reclaim
*crp
= NULL
;
3104 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3105 while (!list_empty(&reclaim_str_hashtbl
[i
])) {
3106 crp
= list_entry(reclaim_str_hashtbl
[i
].next
,
3107 struct nfs4_client_reclaim
, cr_strhash
);
3108 list_del(&crp
->cr_strhash
);
3110 reclaim_str_hashtbl_size
--;
3113 BUG_ON(reclaim_str_hashtbl_size
);
3117 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
3118 static struct nfs4_client_reclaim
*
3119 nfs4_find_reclaim_client(clientid_t
*clid
)
3121 unsigned int strhashval
;
3122 struct nfs4_client
*clp
;
3123 struct nfs4_client_reclaim
*crp
= NULL
;
3126 /* find clientid in conf_id_hashtbl */
3127 clp
= find_confirmed_client(clid
);
3131 dprintk("NFSD: nfs4_find_reclaim_client for %.*s with recdir %s\n",
3132 clp
->cl_name
.len
, clp
->cl_name
.data
,
3135 /* find clp->cl_name in reclaim_str_hashtbl */
3136 strhashval
= clientstr_hashval(clp
->cl_recdir
);
3137 list_for_each_entry(crp
, &reclaim_str_hashtbl
[strhashval
], cr_strhash
) {
3138 if (same_name(crp
->cr_recdir
, clp
->cl_recdir
)) {
3146 * Called from OPEN. Look for clientid in reclaim list.
3149 nfs4_check_open_reclaim(clientid_t
*clid
)
3151 return nfs4_find_reclaim_client(clid
) ? nfs_ok
: nfserr_reclaim_bad
;
3154 /* initialization to perform at module load time: */
3157 nfs4_state_init(void)
3161 status
= nfsd4_init_slabs();
3164 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3165 INIT_LIST_HEAD(&conf_id_hashtbl
[i
]);
3166 INIT_LIST_HEAD(&conf_str_hashtbl
[i
]);
3167 INIT_LIST_HEAD(&unconf_str_hashtbl
[i
]);
3168 INIT_LIST_HEAD(&unconf_id_hashtbl
[i
]);
3170 for (i
= 0; i
< FILE_HASH_SIZE
; i
++) {
3171 INIT_LIST_HEAD(&file_hashtbl
[i
]);
3173 for (i
= 0; i
< OWNER_HASH_SIZE
; i
++) {
3174 INIT_LIST_HEAD(&ownerstr_hashtbl
[i
]);
3175 INIT_LIST_HEAD(&ownerid_hashtbl
[i
]);
3177 for (i
= 0; i
< STATEID_HASH_SIZE
; i
++) {
3178 INIT_LIST_HEAD(&stateid_hashtbl
[i
]);
3179 INIT_LIST_HEAD(&lockstateid_hashtbl
[i
]);
3181 for (i
= 0; i
< LOCK_HASH_SIZE
; i
++) {
3182 INIT_LIST_HEAD(&lock_ownerid_hashtbl
[i
]);
3183 INIT_LIST_HEAD(&lock_ownerstr_hashtbl
[i
]);
3185 memset(&onestateid
, ~0, sizeof(stateid_t
));
3186 INIT_LIST_HEAD(&close_lru
);
3187 INIT_LIST_HEAD(&client_lru
);
3188 INIT_LIST_HEAD(&del_recall_lru
);
3189 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++)
3190 INIT_LIST_HEAD(&reclaim_str_hashtbl
[i
]);
3191 reclaim_str_hashtbl_size
= 0;
3196 nfsd4_load_reboot_recovery_data(void)
3201 nfsd4_init_recdir(user_recovery_dirname
);
3202 status
= nfsd4_recdir_load();
3203 nfs4_unlock_state();
3205 printk("NFSD: Failure reading reboot recovery data\n");
3209 get_nfs4_grace_period(void)
3211 return max(user_lease_time
, lease_time
) * HZ
;
3215 * Since the lifetime of a delegation isn't limited to that of an open, a
3216 * client may quite reasonably hang on to a delegation as long as it has
3217 * the inode cached. This becomes an obvious problem the first time a
3218 * client's inode cache approaches the size of the server's total memory.
3220 * For now we avoid this problem by imposing a hard limit on the number
3221 * of delegations, which varies according to the server's memory size.
3224 set_max_delegations(void)
3227 * Allow at most 4 delegations per megabyte of RAM. Quick
3228 * estimates suggest that in the worst case (where every delegation
3229 * is for a different inode), a delegation could take about 1.5K,
3230 * giving a worst case usage of about 6% of memory.
3232 max_delegations
= nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT
);
3235 /* initialization to perform when the nfsd service is started: */
3238 __nfs4_state_start(void)
3240 unsigned long grace_time
;
3242 boot_time
= get_seconds();
3243 grace_time
= get_nfs4_grace_period();
3244 lease_time
= user_lease_time
;
3245 locks_start_grace(&nfsd4_manager
);
3246 printk(KERN_INFO
"NFSD: starting %ld-second grace period\n",
3248 laundry_wq
= create_singlethread_workqueue("nfsd4");
3249 queue_delayed_work(laundry_wq
, &laundromat_work
, grace_time
);
3250 set_max_delegations();
3254 nfs4_state_start(void)
3258 nfsd4_load_reboot_recovery_data();
3259 __nfs4_state_start();
3265 nfs4_lease_time(void)
3271 __nfs4_state_shutdown(void)
3274 struct nfs4_client
*clp
= NULL
;
3275 struct nfs4_delegation
*dp
= NULL
;
3276 struct list_head
*pos
, *next
, reaplist
;
3278 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++) {
3279 while (!list_empty(&conf_id_hashtbl
[i
])) {
3280 clp
= list_entry(conf_id_hashtbl
[i
].next
, struct nfs4_client
, cl_idhash
);
3283 while (!list_empty(&unconf_str_hashtbl
[i
])) {
3284 clp
= list_entry(unconf_str_hashtbl
[i
].next
, struct nfs4_client
, cl_strhash
);
3288 INIT_LIST_HEAD(&reaplist
);
3289 spin_lock(&recall_lock
);
3290 list_for_each_safe(pos
, next
, &del_recall_lru
) {
3291 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3292 list_move(&dp
->dl_recall_lru
, &reaplist
);
3294 spin_unlock(&recall_lock
);
3295 list_for_each_safe(pos
, next
, &reaplist
) {
3296 dp
= list_entry (pos
, struct nfs4_delegation
, dl_recall_lru
);
3297 list_del_init(&dp
->dl_recall_lru
);
3298 unhash_delegation(dp
);
3301 nfsd4_shutdown_recdir();
3306 nfs4_state_shutdown(void)
3308 cancel_rearming_delayed_workqueue(laundry_wq
, &laundromat_work
);
3309 destroy_workqueue(laundry_wq
);
3310 locks_end_grace(&nfsd4_manager
);
3312 nfs4_release_reclaim();
3313 __nfs4_state_shutdown();
3314 nfs4_unlock_state();
3318 * user_recovery_dirname is protected by the nfsd_mutex since it's only
3319 * accessed when nfsd is starting.
3322 nfs4_set_recdir(char *recdir
)
3324 strcpy(user_recovery_dirname
, recdir
);
3328 * Change the NFSv4 recovery directory to recdir.
3331 nfs4_reset_recoverydir(char *recdir
)
3336 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
3340 if (S_ISDIR(path
.dentry
->d_inode
->i_mode
)) {
3341 nfs4_set_recdir(recdir
);
3349 nfs4_recoverydir(void)
3351 return user_recovery_dirname
;
3355 * Called when leasetime is changed.
3357 * The only way the protocol gives us to handle on-the-fly lease changes is to
3358 * simulate a reboot. Instead of doing that, we just wait till the next time
3359 * we start to register any changes in lease time. If the administrator
3360 * really wants to change the lease time *now*, they can go ahead and bring
3361 * nfsd down and then back up again after changing the lease time.
3363 * user_lease_time is protected by nfsd_mutex since it's only really accessed
3364 * when nfsd is starting
3367 nfs4_reset_lease(time_t leasetime
)
3369 user_lease_time
= leasetime
;